52 comments

  • lbreakjai12 hours ago
    I consider myself rather smart and good at what I do. It's nice to have a look at problems like these once in a while, to remind myself of how little I know, and how much closer I am to the average than to the top.
    • epolanski3 hours ago
      Computing is a very broad topic. Even Linus or Carmack have no skills or knowledge about countless topics that would be mundane to you.<p>It doesn&#x27;t matter really, what matters is our ability to stare into the void of what we don&#x27;t know and start making progress.<p>Our ability to process and master new topics is part of the job.<p>I&#x27;m sure you&#x27;ve done that countless times.
    • TrackerFF11 hours ago
      Well it is a specialized problem. If you&#x27;ve never worked on anything similar previously, it is going to take time. Don&#x27;t even need to interview for selective billion dollar companies like Anthropic to encounter these types of problems - after college I interviewed for various electronics&#x2F;hardware companies where you&#x27;d get asked to optimize low-level code - which would have looked quite foreign, if you had never actually worked on such problems before.
      • Onavo10 hours ago
        If you ask an EE to debug react state management code without prior exposure they won&#x27;t do too well either. But on the other hand they can easily pick up most of it after a week long crash course while training a performance engineer who can optimize code for a specific architecture would take months.
        • sublinear4 hours ago
          &gt; they can easily pick up most of it after a week long crash course<p>I have to disagree and question what you mean by &quot;optimization&quot;. It&#x27;s very easy to write web code that technically accomplishes a task, but does so poorly. This is the natural consequence of having so many options available.<p>The vast majority of web devs with less than 5 years of experience simply don&#x27;t understand plain javascript well enough. It&#x27;s a longstanding problem that devs will reach for the most ergonomic tools, not the best tools.<p>Lacking sufficient experience, they can&#x27;t help it. This happens in all programming languages and in all layers of software. AI slop is even worse because it tends towards the mean.
          • ontouchstart3 hours ago
            Engineering is more or less about getting familiar with the proper tools and use them to solve specific problems: add new features, debugging, refactoring and optimizing.<p>And the tools themselves are built by other engineers and they need new features, debugging, optimization etc. It is turtles all the way down.<p>But each layer has its own jargons, conventions and unwritten hacks. That is where experience comes in. Once you get out off a rabbit hole or pothole, you are one step closer to becoming the “domain expert”. There is no short cut.
        • ignoramous9 hours ago
          &gt; <i>EE to debug react state management ... easily pick up most of it after a week long crash course while training a performance engineer ... would take months</i><p>Isn&#x27;t that mostly because as you go up the abstraction layer, tools and docs to teach yourself the tricks of trade <i>fast</i> are in abundance (let alone a popular layer like React)? Which inturn is likely a function of incentives and opportunities.
          • fny8 hours ago
            It&#x27;s because the higher up the stack you go, tools become more declarative and literate. Calling sort is far easier than understanding the algorithm for example.
            • giancarlostoro5 hours ago
              &gt; Calling sort is far easier than understanding the algorithm for example.<p>This was one of my gripes in college, why am I implementing something if I just need to understand what it does? I&#x27;m going to use the built-in version anyway.
              • Fwirt1 minute ago
                My high school computer science teacher (best one I ever had) once told us this anecdote when we were learning sorting algorithms:<p>He was brought in by the state to do some coaching for existing software devs back in the 90s. When he was going over the various different basic algorithms (insertion sort, selection sort, etc.) one of the devs in the back of the class piped up with, &quot;why are you wasting our time? C++ has qsort built in.&quot;<p>When you&#x27;re processing millions of records, many of which are probably already sorted, using an insertion sort to put a few new records into a sorted list, or using selection sort to grab the few records you need to the front of the queue, is going to be an order of magnitude faster than just calling qsort every time.<p>Turned out he worked for department of revenue. So my teacher roasted him with &quot;oh, so you&#x27;re the reason it takes us so long to get our tax returns back.&quot;<p>Thinking that you can just scoot by using the built-in version is how we get to the horrible state of optimization that we&#x27;re in. Software has gotten slow because devs have gotten lazy and don&#x27;t bother to understand the basics of programming anymore. We should be running a machine shop, not trying to build a jet engine out of Lego.
              • robmccoll4 hours ago
                Because that&#x27;s the entire point of college. It&#x27;s supposed to teach you the fundamentals - how to think, how to problem solve, how to form mental models and adapt them, how things you use actually work. Knowing how different sorting functions work and what the tradeoffs are allows you to pick the best sorting function for your data and hardware. If the tools you have aren&#x27;t doing the job, you can mend them or build new tools.
              • godelski3 hours ago
                So you know which sort to call because there isn&#x27;t a right answer for all cases.<p>And so you can write your own because you&#x27;re probably going to want to sort data in a specific way. Sort doesn&#x27;t mean in numerical increasing or decreasing order, it means whatever order you want. You&#x27;re sorting far more often than you&#x27;re calling the sort function.
              • ksenzee4 hours ago
                The problem is that a computer science degree isn&#x27;t the right training for most software engineering jobs.
              • komali25 hours ago
                &gt; why am I implementing something if I just need to understand what it does?<p>So you can pass job interviews, of course!
    • fergie11 hours ago
      I&#x27;m 30 years in, and literally don&#x27;t understand the question.
      • WithinReason8 hours ago
        After a quick look this is can be seen as a low level GPU&#x2F;TPU optimization problem where you have to consider the throughput and depth of different arithmetic pipelines. If you want to hire people who understand how to do that you unfortunately have to give them such a convoluted task and emulate the relevant parts of HW. (In reality this is probably more like TPU since it has scalar pipelines, but the optimization methods are not that different)<p>The task is to parallelize tree traversal, which is embarrassingly unparallel so it&#x27;s tricky.
        • WithinReason5 hours ago
          This also shows that a performance engineer&#x27;s job, even at Anthropic, is to be a glorified human compiler, who is often easily beaten by LLMs.
          • scottyah1 hour ago
            I think the job is to be one of the few that&#x27;s better than LLMs.
      • mike_hearn10 hours ago
        The question isn&#x27;t clearly written down anywhere, that&#x27;s why. Presumably actual candidates would have been given more info over the phone or email. Part of the &quot;challenge&quot; is reverse engineering their Python; unclear if that&#x27;s intentional.<p>If you look at the top of perf_takehome.py then there is a brief comment saying the challenge is to optimize a kernel. Kernel in GPU land means a program that computes on data in parallel, it&#x27;s not an OS kernel:<p><pre><code> Optimize the kernel (in KernelBuilder.build_kernel) as much as possible in the available time, as measured by test_kernel_cycles on a frozen separate copy of the simulator. </code></pre> However, this kernel doesn&#x27;t run on an actual GPU. It runs on a little interpreter for a custom assembly language written in Python. Thus you will be optimizing the program built in-memory by the function on this line:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;blob&#x2F;f88c9458dbc5ef09d7801d961d905698e7e7cae1&#x2F;perf_takehome.py#L85" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;...</a><p>This function is described only as:<p><pre><code> Like reference_kernel2 but building actual instructions. Scalar implementation using only scalar ALU and load&#x2F;store. </code></pre> The KernelBuilder class has some fields like &quot;instrs&quot; but we can&#x27;t immediately see what they&#x27;re meant to be because this is Python and types are optional. Nonetheless we can see that instructions are being added to a list, and below we can see the test_kernel_cycles function that runs the interpreter on the program. So our mission is to change the build_kernel function to make a better program. And it says this is an assembly version of the python function reference_kernel2 which is found in problem.py.<p>What exactly is this kernel doing? The reference_kernel2 function doesn&#x27;t explain itself either - it&#x27;s some sort of parallel tree walk. Let&#x27;s put that to one side for a second and explore the machine, which is defined in problem.py. The machine itself is also largely undocumented, but there&#x27;s a brief description in a docstring on line 66.<p>At this point it helps to understand the design of exotic processors. The emulator is for a fictional CPU that uses a VLIW SIMD ISA. Normal programmers will never encounter such a chip. Intel tried to make such a machine decades ago and it never took off, since then the concept has been largely dead. I believe it&#x27;s still used in some mobile DSPs like Qualcomm&#x27;s Hexagon. Notably, NVIDIA PTX is <i>not</i> such an ISA so this seems to have been chosen just to make things harder. As the comment explains, in a VLIW machine multiple instructions are packed together into a &quot;slot&quot; and executed in parallel. In a normal CPU the hardware reads a serial stream of instructions and works out just in time which can be executed in parallel, using fancy out-of-order circuitry. In a VLIW machine that&#x27;s done ahead of time by the compiler or (in this case) the humble programmer, you. But this isn&#x27;t just a VLIW machine, it&#x27;s also multi-core, and multi-&quot;engine&quot;, so there are multiple levels of execution going on. And it&#x27;s SIMD, meaning each instruction can itself operate on multiple bits of data simultaneously.<p>This machine doesn&#x27;t have registers or cache but it does have &quot;scratch space&quot;, and so you can use the vector instructions to load data into a series of 32 bit scratch words and then do things on them in parallel. And multiple vector instructions can also run in parallel. &quot;Broadcasting a scalar&quot; in SIMD-speak means taking a single value and repeating it over multiple scratch space slots (or register subwords in a real machine), so you take e.g. 0xFF and get 0xFFFFFFFFFFFFFFFF.<p>And that&#x27;s it, that&#x27;s all we get. As the code says: &quot;This comment is not meant to be full ISA documentation though, for the rest you should look through the simulator code&quot;. Possible point of confusion: real ISAs are serialized to bytes but this one is just Python tuples. The code is only partially typed; sometimes you&#x27;re just left guessing.<p>So to recap, the problem is to optimize an undocumented program expressed in undocumented data structures returned by a Python function whose result is interpreted by a partly documented Python class that simulates a fictional exotic CPU architecture using an abandoned design that gives a lot of parallel computational capacity, but which requires all parallelism to be statically declared ahead of time, whilst simultaneously reverse engineering the Python that does all this.<p>Does that help? Sounds like a fun exercise :)<p><i>Edit: I just checked and Google TPUs are much more VLIW like so perhaps this simulator is designed to match a TPU. I know Anthropic rely on TPUs for serving and have done some optimization for them.</i>
        • HarHarVeryFunny5 hours ago
          It does seem a bit of a strange challenge - a bit reminiscent of high school math problems where understanding the question was as much part of it as actually solving the problem when you understood it.<p>Since the focus of the challenge appears(?) intended to be optimization, not reverse engineering, it&#x27;s a bit odd that they don&#x27;t give a clear statement of what the kernel is meant to be computing. Perhaps the challenge is intended to be a combination of the two, but then the correct reverse engineering part of it becomes a gate for the optimization part, else you&#x27;ll be solving the wrong problem.<p>Given the focus on results achieved by Opus 4.5, maybe that&#x27;s the main point - to show how well Opus can reverse engineer something like this. If they gave the actual clear problem statement, then maybe you could brute force an optimal solution using tree search.
          • fc417fc8022 hours ago
            This isn&#x27;t &quot;reverse engineering&quot; it&#x27;s merely &quot;being able to read fairly simple code you didn&#x27;t write&quot;. A much simpler version of the kernel is provided at the end of problem.py as reference_kernel2.<p>If you can&#x27;t make sense of such a small codebase or don&#x27;t immediately recognize the algorithm that&#x27;s being used (I&#x27;m guilty of the latter) then you presumably aren&#x27;t someone that they want to hire.
            • HarHarVeryFunny2 hours ago
              Fair enough, and there are clues in the comments too, but why not just provide the specification of the kernel (inputs and outputs) as part of the problem?
              • fc417fc8021 hour ago
                They do. They provide reference_kernel which shows the algorithm itself, build_mem_image which shows the data format you will be working with, and finally reference_kernel2 which implements said algorithm on said data format.<p>They then provide you with a very naive implementation that runs on their (very simple) VLIW architecture that you are to optimize.<p>If at the end of that someone is still lost I think it is safe to say it was their goal that person should fail.
                • HarHarVeryFunny1 hour ago
                  Well, yes, they have a reference implementation as documentation, just as they have the simulator as documentation for the ISA ...<p>The problem is about pipelining memory loads and ALU operations, so why not just give clear documentatation and state the task rather than &quot;here&#x27;s a kernel - optimize it&quot;? \_(ツ)_&#x2F;
                  • fc417fc80249 minutes ago
                    Presumably that is only one of two purposes, with the other being to test your ability to efficiently read, understand, and edit low level code that you didn&#x27;t write. I imagine you&#x27;d regularly run into raw PTX if you worked for them in the relevant capacity.<p>And perhaps a third purpose is to use the simulator to test your ability to reason about hardware that you are only just getting familiar with.
          • HarHarVeryFunny4 hours ago
            I just threw this prompt at Gemini, and it seems (I haven&#x27;t analyzed the problem to see if it is correct), to be able to extract a clear understanding of the problem, and a specification for the kernel.<p>&quot;Can you &quot;reverse engineer&quot; what the kernel in this optimization exercise is actually doing - write a specification for it?<p><a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome</a>&quot;<p>Gemini says it&#x27;s doing inference on a random forest - taking a batch of inputs, running each one through each decision tree, and for each input outputting the sum of these decision tree outputs - the accumulated evidence.
            • HarHarVeryFunny17 minutes ago
              So looking at the actual code (reference_kernel() in problem.py), this &quot;random forest inference&quot; is completely wrong!<p>It&#x27;s doing some sort of binary tree traversal, but the hashing and wrap around looks weird - maybe just a made up task rather than any useful algorithm?
        • owlbite4 hours ago
          I think calling VLIW &quot;an adandoned design&quot; is somewhat of an exaggeration, such architectures are pretty common for embedded audio processing.
          • matt_d1 hour ago
            Worth adding on that note:<p>From JAX to VLIW: Tracing a Computation Through the TPU Compiler Stack, <a href="https:&#x2F;&#x2F;patricktoulme.substack.com&#x2F;p&#x2F;from-jax-to-vliw-tracing-a-computation" rel="nofollow">https:&#x2F;&#x2F;patricktoulme.substack.com&#x2F;p&#x2F;from-jax-to-vliw-tracin...</a><p>Google’s Training Chips Revealed: TPUv2 and TPUv3, HotChips 2020, <a href="https:&#x2F;&#x2F;hc32.hotchips.org&#x2F;assets&#x2F;program&#x2F;conference&#x2F;day2&#x2F;HotChips2020_ML_Training_Google_Norrie_Patil.v01.pdf" rel="nofollow">https:&#x2F;&#x2F;hc32.hotchips.org&#x2F;assets&#x2F;program&#x2F;conference&#x2F;day2&#x2F;Hot...</a><p>Ten Lessons From Three Generations Shaped Google’s TPUv4i, ISCA 2021, <a href="https:&#x2F;&#x2F;gwern.net&#x2F;doc&#x2F;ai&#x2F;scaling&#x2F;hardware&#x2F;2021-jouppi.pdf" rel="nofollow">https:&#x2F;&#x2F;gwern.net&#x2F;doc&#x2F;ai&#x2F;scaling&#x2F;hardware&#x2F;2021-jouppi.pdf</a>
          • HarHarVeryFunny1 hour ago
            x86-64 SSE and AVX are also SIMD
          • mike_hearn4 hours ago
            Sure. I did mention DSPs. But how many people write code for DSPs?
        • forgotpwd1610 hours ago
          This is nice writeup. Thanks. Another commenter said will&#x27;ve taken them 2h just to sketch out ideas; sans LLMs will&#x27;ve taken me more than 2h just to collect all this info let alone start optimizing it.
          • mike_hearn9 hours ago
            It took me about 10 minutes to generate that writeup the old fashioned 100% organic way, because one of the things that&#x27;s unspecified is whether you&#x27;re allowed to use AI to help solve it! So I assumed as it&#x27;s a job interview question you&#x27;re not allowed, but now I see other comments saying it was allowed. That would let you get much further.<p>I think I&#x27;d be able to make some progress optimizing this program in two hours but probably not much. I&#x27;m not a performance engineer but have designed exotic emulated CPU architectures before, so that helps a lot.
          • maccard7 hours ago
            I&#x27;ve not written a VM before, but the comments in perf_takehome.py and problem.py explain the basics of this.<p>I gleaned about half of this comment in a few minutes of just skimming the code and reading the comments on the functions and classes. There&#x27;s only 500 lines of code really (the rest is the benchmark framework).
            • fc417fc8023 hours ago
              Same thought. I doubt they provided additional explanation to candidates - it seems that basic code literacy within the relevant domain is one of the first things being tested.<p>On the whole I don&#x27;t think I&#x27;d perform all that well on this task given a short time limit but it seems to me to be an extremely well designed task given the stated context. The reference kernel easily fits on a single screen and even the intrinsic version almost does. I think this task would do a good job filtering the people they don&#x27;t want working for them (and it seems quite likely that I&#x27;m borderline or maybe worse by their metric).
        • b40d-48b2-979e6 hours ago
          <p><pre><code> Sounds like a fun exercise :) </code></pre> I&#x27;ll be honest, that sounds like the opposite of fun since the worst parts of my job are touching the parts of a Python codebase that are untyped. The sad part is this work codebase isn&#x27;t even that old, maybe a few years, and the developers definitely should have known better if they had anyone capable leading them. Alas, they&#x27;re all gone now.<p>Harder than figuring out the instruction set for some exotic CPU are definitely the giant untyped dicts&#x2F;lists common in data science code.
        • carschno9 hours ago
          On the one hand, this exercise probably reflects a realistic task. Daily engineering work comprises a lot of reverse engineering and debugging of messy code. On the other hand, this does not seem very suitable as an isolated assignment. The lack of code base-specific context has a lot of potential for frustration. I wonder what they really tested on the candidates, and whether this was what they wanted to filter for.
          • fc417fc8023 hours ago
            &gt; The lack of code base-specific context has a lot of potential for frustration.<p>I think that&#x27;s one of the intentional points. Being able to quickly understand what the provided source code is doing.
        • dist-epoch8 hours ago
          &gt; but which requires all parallelism to be statically declared ahead of time<p>this is what all specialized chips like TPU&#x2F;Cerebras require today, and it allows for better optimization than a generic CPU since you can &quot;waste&quot; 30 min figuring out the perfect routing&#x2F;sequencing of operations, instead of doing it in the CPU in nanoseconds&#x2F;cycles<p>another benefit is you can throw away all the CPU out-of-order&#x2F;branch prediction logic and put useful matrix multipliers in it&#x27;s place
        • mannyv1 hour ago
          &quot;Performance can be optimized by not using python.&quot;
        • fergie8 hours ago
          Wow! Thanks for the explanation :)
      • karmajunkie2 hours ago
        Thank goodness, I thought it was just me...
      • PeterStuer10 hours ago
        Which part exactly are ypu having trouble with?<p>- Optimize the kernel (in KernelBuilder.build_kernel) as much as possible in the available time, as measured by test_kernel_cycles on a frozen separate copy of the simulator
      • bsder8 hours ago
        Since it&#x27;s a CPU, you start with the idea that there is an ALU and spiral outward from that. That gives you something concrete to wrap your head around while you climb up the abstraction levels.<p>However, when I hit &quot;scratch_write&quot; and it wasn&#x27;t in the Machine class and it wasn&#x27;t coming from some Decorator and it was getting defined <i>and deleted</i> by a <i>member function</i> ... I stopped. That&#x27;s paying lip service to the variable typing that is scattered around and actively hampers even basic IDE usage. Probably the typing was added by AI&#x2F;LLM after the fact, and it missed that unusual usage. The Python convention used to be that those kinds of variables got declared as &quot;_scratch_write&quot; with a leading underscore to flag that they were &quot;private&#x2F;internal&quot;.<p>That was the gigantic red &quot;We write shitty code&quot; signal or worse &quot;We don&#x27;t care about wasting your time&quot; signal. Human review should have flagged that.<p>Shame. I was kinda looking forward to the technical problem, but I&#x27;m not going to spend a bunch of time using grep to untangle garbage code to get at it.<p>I suspect everything would actually be much clearer if you wrote it in SystemVerilog and tested with Cocotb. Let&#x27;s see if their LLMs can handle <i>that</i> porting job. HAH!
      • measurablefunc11 hours ago
        Generate instructions for their simulator to compute some numbers (hashes) in whatever is considered the memory of their &quot;machine&quot;¹. I didn&#x27;t see any places where they actually disallow cheating b&#x2F;c it says they only check the final state of the memory² so seems like if you know the final state you could just &quot;load&quot; the final state into memory. The cycle count is supposedly the LLM figuring out the fewest number of instructions to compute the final state but again, it&#x27;s not clear what they&#x27;re actually measuring b&#x2F;c if you know the final state you can cheat &amp; there is no way to tell how they&#x27;re prompting the LLM to avoid the answers leaking into the prompt.<p>¹<a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;blob&#x2F;main&#x2F;problem.py#L64" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;...</a><p>²<a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;blob&#x2F;main&#x2F;problem.py#L567" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;...</a>
        • saagarjha10 hours ago
          Well, they read your code in the actual hiring loop.
          • measurablefunc10 hours ago
            My point still stands. I don&#x27;t know what the LLM is doing so my guess is it&#x27;s cheating unless there is evidence to the contrary.
            • red75prime9 hours ago
              I guess your answer to &quot;Try to run Claude Code on your own &#x27;ill-defined&#x27; problem&quot; would be &quot;I&#x27;m not interested.&quot; Correct? I think we can stop here then.
            • KeplerBoy8 hours ago
              Well that&#x27;s certainly a challenge when you use LLMs for this test driven style of programming.
            • saagarjha10 hours ago
              Why do you assume it’s cheating?
              • measurablefunc8 hours ago
                Because it&#x27;s a well know failure mode of neural networks &amp; scalar valued optimization problems in general: <a href="https:&#x2F;&#x2F;www.nature.com&#x2F;articles&#x2F;s42256-020-00257-z" rel="nofollow">https:&#x2F;&#x2F;www.nature.com&#x2F;articles&#x2F;s42256-020-00257-z</a>
                • saagarjha7 hours ago
                  Again, you can just read the code
                  • measurablefunc2 hours ago
                    You&#x27;re missing the point. There is no evidence to support their claims which means they are more than likely leaking the memory into the LLM prompt &amp; it is cheating by simply loading constants into memory instead of computing anything. This is why formal specifications are used to constrain optimization. Without proof that the code is equivalent you might as well just load constants into memory &amp; claim victory.
                    • fc417fc8022 hours ago
                      &gt; There is no evidence to support their claims<p>Do you make a habit of not presuming even basic competence? You believe that Anthropic left the task running for hours, got a score back, and never bothered to examine the solution? Not even out of curiosity?<p>Also if it was cheating you&#x27;d expect the final score to be unbelievably low. Unless you also suppose that the LLM actively attempted to deceive the human reviewers by adding extra code to burn (approximately the correct number of) cycles.
                      • measurablefunc48 minutes ago
                        This has nothing to do w&#x2F; me &amp; consistently making it a personal problem instead of addressing the claims is a common tactic for people who do not know what it means to present evidence for their claims. Anthropic has not provided the necessary evidence for me to conclude that their LLM is not cheating. I have no opinion on their competence b&#x2F;c that is not what is at issue. They could be incompetent &amp; not notice that their LLM is cheating at their take home exam but I don&#x27;t care about that.
                        • fc417fc80236 minutes ago
                          You are implying that you believe them to be incompetent since otherwise you would not expect evidence in this instance. They also haven&#x27;t provided independent verification of their claims - do you suspect them of lying as well?<p>How do you explain the specific score that was achieved if as you suggest the LLM simply copied the answer directly?
                • red75prime8 hours ago
                  And? Anthropic is not aware of this 2020 paper? The problem is not solvable?
                  • measurablefunc2 hours ago
                    Why are you asking me? Email &amp; ask Anthropic.
    • ActorNightly33 minutes ago
      Yours is a good mentality to have because it creates the emotional drive to learn more, so don&#x27;t lose that. That being said, this isn&#x27;t really that complicated. Its just a matter of taking enough time to look at the code and understand how its structured. I feel like the thing that differentiates developer skill is pretty much being able to do that, specifically in the process of having the model of the program in your head.
    • mangatmodi8 hours ago
      Smart is different than the knowledge. If you learn about these concepts andwork on these problems, then you will be able to solve them.<p>It&#x27;s not about you being average, just a different knowledge set.
    • elzbardico2 hours ago
      There&#x27;s a big chance you&#x27;re falling in a subtle form of imposter syndrome that manifests itself by largely over-estimating the average skill level.<p>But this is good. Staying humble makes you hungrier for learning.
    • chistev8 hours ago
      What we know is a drop, what we don&#x27;t know is an ocean.
    • LouisSayers6 hours ago
      It&#x27;s the type of thing you&#x27;d be exposed to in a computer science degree - operating systems &#x2F; compilers.<p>Always room to learn in software :)
    • deadbabe5 hours ago
      If you think you’re average, you’re not average.
    • xenihn11 hours ago
      It comes with test suites, so that gives you a base to start from. You can at the very least do trial-and-error and come up with some heuristics on the fly. You&#x27;re at a huge disadvantage to someone who has some familiarity but can convincingly play it off as being a newcomer, though.
    • apsurd12 hours ago
      disagree. nobody has a monopoly on what metric makes someone good. I don&#x27;t understand all this leet code optimization. actually i do understand it, but it&#x27;s a game that will attract game optimizers.<p>the hot take is, there are other games.
      • tuetuopay10 hours ago
        This is the opposite of leet code.<p>Yes, this applies to some simulated imaginary CPU with an artificial problem. Except that the job asked here is exactly the core of what a performance engineer will do at anthropic: optimize kernels for their fleet of GPUs. Is it simplified? Yes! (e.g. the simulator does not restrict memory access patterns)<p>This is a real-world problem adapted to a lab setting that can fit in one&#x27;s head in a matter of hours. Leetcode would have you reimplement the hashmap used in there.
      • saagarjha10 hours ago
        This is explicitly not Leetcode, in fact its goal is to attract optimizers
      • sevenzero11 hours ago
        Also leetcode does not really provide insight into ones ability to design business solutions. Whether it be system design, just some small feature implementation or communication skills within a team. Its just optimizers jerking each other off on some cryptic problems 99.999999999% of developers will never see in real life. Maybe it would&#x27;ve been useful like 30 years ago, but all commonly used languages have all these fancy algorithms baked into their stdlib, why would I ever have to implement them myself?
        • lbreakjai11 hours ago
          But this is an interview problem at Anthropic, not at your local CRUD factory. They _are_ looking for the optimizers, because they _are_ working on cryptic problems the 99.9999% of us will never encounter.
        • thorncorona11 hours ago
          Or more likely, the commonality is how you&#x27;re applying your software skills?<p>In every other field it&#x27;s helpful to understand the basics. I don&#x27;t think software is the exception here.
          • sevenzero11 hours ago
            Understanding basics is very different to being able to memorize algorithms. I really dont see why I&#x27;d ever have to implement stuff like quicksort myself somewhere. Yes I know what recursion is, yes I know what quick sort is, so if I ever need it I know what to look for. Which was good enough throughout my career.
  • game_the0ry4 hours ago
    &gt; If you optimize below 1487 cycles, beating Claude Opus 4.5&#x27;s best performance at launch, email us at performance-recruiting@anthropic.com with your code (and ideally a resume) so we can be appropriately impressed and perhaps discuss interviewing.<p>This is an interesting way to recruit. Much better than standard 2 leetcode medium&#x2F;hard questions in 45 mins.
    • paxys3 hours ago
      This is simply to enter the recruiting pipeline. once you&#x27;re in you will do the same leetcode interviews as everyone else.
      • alt2273 hours ago
        You would hope that if you manage to beat their engineers best optimisations at launch, then you would leapfrog a certain amount of the initial stages.<p>Then again, this may just be a way to get free ideas at optimising their product from outside the box.
        • benlivengood2 hours ago
          One could use any number of LLMs on a take-home problem so in-person interviews are a must.
          • legel2 hours ago
            One could use any number of LLMs on real-world problems.<p>Why are we still interviewing like its 1999?
            • game_the0ry2 hours ago
              Old habits die hard. And engineers are pretty lazy when it comes to interviews, so just throwing the same leetcode problem into coder pad in every interview makes interviews easier for the person doing the interview.
              • selkin2 hours ago
                If you want people to interview better, you have to both allocate resources to it, and make it count on perf. It’s not laziness, it’s ROI.
              • yodsanklai2 hours ago
                As an interviewer, I ask the same problems because it makes it much easier to compare candidates.
                • game_the0ry2 hours ago
                  How do you know if one candidate happened to see the problem on leetcode and memorized the solution versus one who struggled but figured it out slower?
                  • yodsanklai1 hour ago
                    It&#x27;s very easy to tell, but it doesn&#x27;t make much difference. The best candidates have seen the problems before and don&#x27;t even try to hide it, they just propose their solution right away.<p>I try give positive feedback for candidates who didn&#x27;t know the problem but could make good use of hints, or had the right approach. But unfortunately, it&#x27;s difficult to pass a Leetcode interview if you haven&#x27;t seen a similar problem to what is asked before. Most candidates I interview nowadays seem to know all questions.<p>That&#x27;s what the company has decided so we have to go along. The positive side is that if you do your part, you have good chances of being hired, even if you disagree with the process.
                  • bradlys2 hours ago
                    It doesn’t matter. It’s about looking for candidates who have put in the time for your stupid hazing ritual. It signals on people who are willing to dedicate a lot of time to meaningless endeavors for the sake of employment.<p>This type of individual is more likely to follow orders and work hard - and most importantly - be like the other employees you hired.
                    • legel54 minutes ago
                      Once upon a time, the &quot;stupid hazing ritual&quot; made sense.<p>Now it means company is stupid.
      • driverdan1 hour ago
        Is this a fact or an assumption?
    • yodsanklai2 hours ago
      It would take something like one week full time to work on this. It&#x27;s not something you can do if you have a full-time job and apply to several other companies. I find it unreasonable to ask a candidate to spend that much time for an uncertain result.<p>It&#x27;s true that being ready for leetcode takes practice, but at least it&#x27;s standard so you can re-use the skills to other interviews. Optimizing some generated code is certainly fun, but it&#x27;s as useless as leetcode for your average programmer.
      • tcoff911 hour ago
        As long as there are qualified candidates willing to do unreasonable tasks for the chance to work at a company, there&#x27;s not much incentive for the company to change their system. Those people will also probably work unreasonably hard and make unreasonable sacrifices for the company.
  • pvalue00514 hours ago
    I suspect this was released by Anthropic as a DDOS attack on other AI companies. I prompted &#x27;how do we solve this challenge?&#x27; into gemini cli in a cloned repo and it&#x27;s been running non-stop for 20 minutes :)
    • bjackman11 hours ago
      Lately with Gemini CLI &#x2F; Jules it doesn&#x27;t seem like time spent is a good proxy for difficulty. It has a big problem with getting into loops of &quot;I am preparing the response for the user. I am done. I will output the answer. I am confident. Etc etc&quot;.<p>I see this directly in Gemini CLI as the harness detects loops and bails the reasoning. But I&#x27;ve also just occasionally seen it take 15m+ to do trivial stuff and I suspect that&#x27;s a symptom of a similar issue.
      • aiiotnoodle6 hours ago
        I&#x27;ve noticed using antigravity and vscode, Gemini 3 pro often comes back with model too busy or something like that and basically 500s.<p>Seems like capacity because it works a lot better late at night.<p>I don&#x27;t see the same with the claude models in antigravity.
        • menaerus4 hours ago
          I also noticed that and I also noticed that it starts to struggle when the workspace &quot;tab&quot; you&#x27;re working in gets longer - it basically gets stuck at &quot;Starting agent ...&quot;. I initially thought it must be a very big context that the model is struggling with but since since restarting the &quot;app&quot; and kill -9 fixes it, it suggests that it&#x27;s a local issue. Strange.
        • trillic3 hours ago
          Anecdotally, I notice better performance and output quality across most providers outside of 8a-5p ET.
      • mixel8 hours ago
        I saw this too. Sometimes it &quot;think&quot; inside of the actual output and its much more likely to end up in the loop of &quot;I am ready to answer&quot; while it is doing that already
      • sva_8 hours ago
        I feel like sometimes it just loops those messages when it doesn&#x27;t actually generate new tokens. But I might be wrong
        • bjackman8 hours ago
          There are some other failure modes that all feel kinda vaguely related that probably help with building a hypothesis about what&#x27;s going wrong:<p>Sometimes Gemini tools will just randomly stop and pass the buck back to you. The last thing will be like &quot;I will read the &lt;blah&gt; code to understand &lt;blah&gt;&quot; and then it waits for another prompt. So I just type &quot;continue&quot; and it starts work again.<p>And, sometimes it will spit out the internal CoT directly instead of the text that&#x27;s actually supposed to be user-visible. So sometimes I&#x27;ll see a bunch of paragraphs starting with &quot;Wait, &quot; as it works stuff out and then at the end it says &quot;I understand the issue&quot; or whatever, then it waits for a prompt. I type &quot;summarise&quot; and it gives me the bit I actually wanted.<p>It feels like all these things are related and probably have to do with the higher-level orchestration of the product. Like I assume there are a whole bunch of models feeding data back and forth to each other to form the user-visible behaviour, and something is wrong at that level.
          • hackpelican32 minutes ago
            At one point it started spitting out its CoT in the comments of the code it’s supposed to be changing.
    • bird086112 hours ago
      Which Gemini model did you use? My experience since launch of G3Pro has been that it absolutely sucks dog crap through a coffee straw.
      • pvalue00511 hours ago
        &#x2F;model: Auto (Gemini 3) Let Gemini CLI decide the best model for the task: gemini-3-pro, gemini-3-flash<p>After ~40 minutes, it got to:<p>The final result is 2799 cycles, a 52x speedup over the baseline. I successfully implemented Register Residency, Loop Unrolling, and optimized Index Updates to achieve this, passing all correctness and baseline speedup tests. While I didn&#x27;t beat the Opus benchmarks due to the complexity of Broadcast Optimization hazards, the performance gain is substantial.<p>It&#x27;s impressive as I definitely won&#x27;t be able to do what it did. I don&#x27;t know most of the optimization techniques it listed there.<p>I think it&#x27;s over. I can&#x27;t compete with coding agents now. Fortunately I&#x27;ve saved enough to buy some 10 acre farm in Oregon and start learning to grow some veggies and raise chickens.
        • ece2 minutes ago
          After an hour with a few prompts, the first working version got to 3529 cycles (41x speedup) for me. I was using Gemini 3 pro preview.
        • light_hue_15 hours ago
          Keep in mind that the boat on competing with machines to generate assembly sailed for 99% of programmers half a century ago. It is not surprising that this is an area where AI is strong.
        • IsTom7 hours ago
          Did you check that it did the things it claims it did?
        • triyambakam3 hours ago
          &gt; grow some veggies and raise chickens.<p>Maybe Claude will be able to do that soon, too.
        • apsurd10 hours ago
          we&#x27;ve lost the plot.<p>you can&#x27;t compete with an AI on doing an AI performance benchmark?
          • kqr9 hours ago
            This is not an AI performance benchmark, this is an actual exercise given to potential human employees during a recruitment process.
      • Mashimo12 hours ago
        &gt; sucks dog crap through a coffee straw.<p>That would be impressive.
        • stronglikedan4 hours ago
          Only if the dog didn&#x27;t get too much human food the night before.
        • anematode12 hours ago
          New LLM benchmark incoming? I bet once it&#x27;s done, people will <i>still</i> say it&#x27;s not AGI.
          • dotancohen12 hours ago
            When they get the hardware capable of that, a different industry will be threatened by AI. The oldest industry.
            • darepublic5 hours ago
              Song of Solomon I guess
            • cess1111 hours ago
              Textile?
              • nineteen99910 hours ago
                The emperor&#x27;s (empresses?) new textile.
  • languid-photic12 hours ago
    Naively tested a set of agents on this task.<p>Each ran the same spec headlessly in their native harness (one shot).<p>Results:<p><pre><code> Agent Cycles Time ───────────────────────────────────────────── gpt-5-2 2,124 16m claude-opus-4-5-20251101 4,973 1h 2m gpt-5-1-codex-max-xhigh 5,402 34m gpt-5-codex 5,486 7m gpt-5-1-codex 12,453 8m gpt-5-2-codex 12,905 6m gpt-5-1-codex-mini 17,480 7m claude-sonnet-4-5-20250929 21,054 10m claude-haiku-4-5-20251001 147,734 9m gemini-3-pro-preview 147,734 3m gpt-5-2-codex-xhigh 147,734 25m gpt-5-2-xhigh 147,734 34m </code></pre> Clearly none beat Anthropic&#x27;s target, but gpt-5-2 did slightly better in much less time than &quot;Claude Opus 4 after many hours in the test-time compute harness&quot;.
    • lawrencechen11 hours ago
      codex cli + gpt-5-2-codex-xhigh got to 1606 with the prompt &quot;beat 1487 cycles. go.&quot; ~53 minutes.
      • jstummbillig11 hours ago
        Will you look at this man&#x27;s prompting skills?!
      • dudewhocodes7 hours ago
        Serious prompt engineering right here
      • mettamage7 hours ago
        Wow, is gpt-5-2-codex-xhigh really that good in general? Is this the 200$ per month version?
        • woadwarrior016 hours ago
          gpt-5.2-codex xhigh with OpenAI codex on the $20&#x2F;month plan got to 1526 cycles with OP&#x27;s prompt for me. Meanwhile claude code with Opus 4.5 on the team premium plan ($150&#x2F;month) gave up with a bunch of contrived excuses at 3433 cycles.
    • ponyous12 hours ago
      Very interesting thanks! I wonder what would happen if you kept running Gemini in a loop for a while. Considering how much faster it ended it seems like there is a lot more potential.
    • a24j10 hours ago
      Can you share the agent-comparison harness code or point to something similar? I want to learn about benchmarking models in a basic or practical sense.
      • languid-photic5 hours ago
        Sure!<p><a href="https:&#x2F;&#x2F;github.com&#x2F;voratiq&#x2F;voratiq" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;voratiq&#x2F;voratiq</a>
    • raphaelj8 hours ago
      Could you try with some open-weighted models, e.g. Qwen3-coder, GLM-4.7 or Devstral-2?
    • forgotpwd1612 hours ago
      Could you make a repo with solutions given by each model inside a dir&#x2F;branch for comparison?
      • kitrak9511 hours ago
        Are you giving instructions to a stranger on the internet?
        • forgotpwd1610 hours ago
          Instructions?! Just asked since GP already did it. No need to realize top comment&#x27;s &quot;DDOS attack on other AI companies&quot; joke.
        • edf1311 hours ago
          I think he’s asking rather than giving instructions
    • giancarlostoro12 hours ago
      I do wonder how Grok would compare, specifically their Claude Code Fast model.
  • Cornbilly7 minutes ago
    All this does for me is make me wish that the parallel computing class at my university wasn&#x27;t run so poorly.
  • abra08 hours ago
    This is a really fun problem! I suggest anyone who likes optimization in a very broad sense to try their hand at it. Might be the most fun I&#x27;ve had while interviewing. I had to spend a week-worth of evenings on it to fully scratch the itch, and I managed to get 1112 cycles. But that was mostly manual, before the current crop of agentic models (clopus 4.5, gpt5.2). I wonder how far you can RalphWiggum it!
    • lukah6 hours ago
      I&#x27;ve never heard AI-assisted coding referred to as &quot;RalphWiggum&quot;ing a problem, and now I will have to use that always. Thank you.
      • usgroup6 hours ago
        <a href="https:&#x2F;&#x2F;awesomeclaude.ai&#x2F;ralph-wiggum" rel="nofollow">https:&#x2F;&#x2F;awesomeclaude.ai&#x2F;ralph-wiggum</a>
  • fabian44 hours ago
    [flagged]
    • tap127834874 hours ago
      [flagged]
      • epiccoleman4 hours ago
        It definitely bears all the LLM hallmarks we&#x27;ve come to know. emdash, the &quot;this isn&#x27;t X. it&#x27;s Y&quot; structure - and then, to cap it off, a single pithy sentence to end it.
        • nostrademons4 hours ago
          Also bears all the hallmarks of an ordinary post (by someone fairly educated) on the Internet. This would make sense, because LLMs were trained on lots of ordinary posts on the Internet, plus a fair number of textbooks and scientific papers.
        • haliskerbas4 hours ago
          I&#x27;ve noticed people who are using LLMs more, myself included, are starting to talk like that.<p>Oops I mean, you&#x27;re absolutely right, those ARE hallmark signs of an LLM. Let me breakdown why this isn&#x27;t just your imagination but actually...
  • avaer14 hours ago
    It&#x27;s pretty interesting how close this assignment looks to demoscene [1] golf [2].<p>[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Demoscene" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Demoscene</a> [2] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Code_golf" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Code_golf</a><p>It even uses Chrome tracing tools for profiling, which is pretty cool: <a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;blob&#x2F;main&#x2F;problem.py#L154" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;...</a>
    • wiz21c10 hours ago
      I was in the demoscene long ago and that kind of optimisation is definitely in the ballpark of what we did: optimize algorithm down to machine code level (and additionally, cheat like hell to make you believe we ran the algorithm for real :-)).<p>But to be honest, I wonder what algorithm they implement. I have read the code for 2 minutes, and it sound like random forest prediction. Anyone knows what the code does ?
      • saagarjha10 hours ago
        It’s some useless problem like a random tree walk or something like that, the actual algorithm is not particularly important to the problem
        • psb2178 hours ago
          Yeah, I assume it was partly chosen since the problem structure provides some convenient hooks for selectively introducing subtle and less subtle inefficiencies in the baseline algorithm that match common optimization patterns.
    • KeplerBoy11 hours ago
      perfetto is pretty widely used for such traces, because building a viewer for your traces is a completely avoidable pain.
    • nice_byte14 hours ago
      it&#x27;s designed to select for people who can be trusted to manually write ptx :-)
  • nine_k2 hours ago
    This is a kind of task that&#x27;s best solved by possibly spending more than the allocated 2 hours on it, once any obvious low-hanging fruit is picked. An optimization task is what a machine does best. So the real problem would be to construct a machine that would be able to run the optimization. A right optimization framework that results from the effort could also efficiently solve many more similar problems in the future.<p>I understand that this test is intended to somehow test the raw brianpower, the ability to tackle an unfamiliar and complicated domain, and to work under stress. But I hope it&#x27;s <i>not</i> representative of the actual working conditions at Anthropic. It&#x27;s like asking a candidate to play a Quake deathmatch when hiring to a special forces assault squad.
  • seamossfet53 minutes ago
    I&#x27;m getting flashbacks from my computer engineering curriculum. Probably the first place I&#x27;d start is replacing comparison operators on the ALU with binary arithmetic since it&#x27;s much faster than branch logic. Next would probably be changing the `step` function from brute iterators on the instructions to something closer to a Btree? Then maybe a sparse set for the memory management if we&#x27;re going to do a lot of iterations over the flat memory like this.
  • sureglymop14 hours ago
    Having recently learned more about SIMD, PTX and optimization techniques, this is a nice little challenge to learn even more.<p>As a take home assignment though I would have failed as I would have probably taken 2 hours to just sketch out ideas and more on my tablet while reading the code before even changing it.
    • forgotpwd1611 hours ago
      Unless misread, 2 hours isn&#x27;t the time limit for the candidate to do this but the time Claude eventually needed to outperform best returned solution. Best candidate could&#x27;ve taken 6h~2d to achieve this result.
      • fhd211 hours ago
        Their Readme.md is weirdly obsessed with &quot;2 hours&quot;:<p>&quot;before Claude Opus 4.5 started doing better than humans given only 2 hours&quot;<p>&quot;Claude Opus 4.5 in a casual Claude Code session, approximately matching the best human performance in 2 hours&quot;<p>&quot;Claude Opus 4.5 after 2 hours in our test-time compute harness&quot;<p>&quot;Claude Sonnet 4.5 after many more than 2 hours of test-time compute&quot;<p>So that does make one wonder where this comes from. Could just be LLM generated with a talking point of &quot;2 hours&quot;, models can fall in love with that kind of stuff. &quot;after many more than 2 hours&quot; is a bit of a tell.<p>Would be quite curious to know though. How I usually design take home assignments is:<p>1. Candidate has several _days_ to complete (usually around a week).<p>2. I design the task to only _take_ 2-4 hours, informing the candidate about that, but that doesn&#x27;t mean they can&#x27;t take longer. The subsequent interview usually reveals if they went overboard or struggled more than expected.<p>But I can easily picture some places sending a candidate the assignment and asking them to hand in their work within two hours. Similar to good old coding competitions.
      • alcasa11 hours ago
        No the 2 hours is their time limit for candidates. The thing is that you are allowed to use any non-human help for their take homes (open book), so if AI can solve it in below 2 hours, it&#x27;s not very good at assessing the human.
        • saagarjha10 hours ago
          4 hours but AI help is (was?) allowed. I assume it was retired because of Opus basically oneshotting it
          • alcasa7 hours ago
            Fair enough. I feel like designing AI-proof take-homes is getting ever more futile. Given the questions need to be sufficiently low context to be human-doable in a short time and timespans for AI tasks increasing, I&#x27;m not sure take homes can actually serve any filtering function whatsoever, besides checking if applicants are willing to put in a minimal amount of effort.
  • bytesandbits12 hours ago
    Having done a bunch of take home for big (and small) AI labs during interviews, this is the 2nd most interesting one I have seen so far.
    • petters12 hours ago
      And the answer to the obvious follow-up question is...?
      • mrklol10 hours ago
        Milk before cereals
        • matthews38 hours ago
          Milk, then cereal, then bowl!
          • Xmd5a8 hours ago
            How about a bowl, and then, 30 minutes ~ 1 hour later, milk with cereals?
      • kevthecoder8 hours ago
        42
      • darkwater10 hours ago
        Maybe it&#x27;s under NDA :)
      • reader927411 hours ago
        fries
  • amirhirsch1 hour ago
    I&#x27;m at 1137 with one hour with opus now... Pipelined vectorized hash, speculation, static code for each stage, epilogues and prologues for each stage-to-stage...<p>I think I&#x27;m going to get sub 900 since i just realized i can in-parallel compute whether stage 5 of the hash is odd just by looking at bits 16 and 0 of stage 4 with less delay.....
    • lalaland11251 hour ago
      How do you avoid the load bottleneck?
      • amirhirsch1 hour ago
        take advantage of index collisions, optimizing round 0 and 11, speculative pre-loading, and the early branch predictor (which now I am doing looking at bits output at stage 3)
  • koolba15 hours ago
    What is the actual assignment here?<p>The README only gives numbers without any information on what you’re supposed to do or how you are rated.
    • glalonde15 hours ago
      &quot;Optimize the kernel (in KernelBuilder.build_kernel) as much as possible in the available time, as measured by test_kernel_cycles on a frozen separate copy of the simulator.&quot; from perf_takehome.py
    • vermilingua14 hours ago
      Think that means you failed :(
      • nice_byte14 hours ago
        +1<p>being cryptic and poorly specified is part of the assignment<p>just like real code<p>in fact, it&#x27;s _still_ better documented an self contained than most of the problems you&#x27;d usually encounter in the wild. pulling on a thread to end up with a clear picture of what needs to be accomplished is like 90% of the job very often.
        • throwaway8152313 hours ago
          I didn&#x27;t see much cryptic except having to click on &quot;perf_takehome.py&quot; without being told to. But, 2 hours didn&#x27;t seem like much to bring the sample code into some kind of test environment, debug it enough to works out details of its behaviour, read through the reference kernel and get some idea of what the algorithm is doing, read through the simulator to understand the VM instruction set, understand the test harness enough to see how the parallelism works, re-code the algorithm in the VM&#x27;s machine language while iterating performance tweaks and running simulations, etc.<p>Basically it&#x27;s a long enough problem that I&#x27;d be annoyed at being asked to do it at home for free, if what I wanted from that was a shot at an interview. If I had time on my hands though, it&#x27;s something I could see trying for fun.
          • ithkuil11 hours ago
            My instinct to read about the problem was to open the &quot;problem.py&quot; file, which states &quot;Read the top of perf_takehome.py for more introduction&quot;<p>So yeah. They _could_ have written it much more clearly in the readme.
          • tayo4213 hours ago
            2 hours does seem short. It took me a half hour to get through all you listed and figure out how to get the valu instruction working.<p>I suspect it would take me another hour to get it implemented. Leaving 30 minutes to figure out something clever?<p>Idk maybe I&#x27;m slow or really not qualified.
          • nice_byte13 hours ago
            it&#x27;s &quot;cryptic&quot; for an interview problem. e.g. the fact that you have to actually look at the vm implementation instead of having the full documentation of the instruction set from the get go.
            • throwaway8152312 hours ago
              That seems normal for an interview problem. They put you in front of some already-written code and you have to fix a bug or implement a feature. I&#x27;ve done tons of those in live interviews. So that part didn&#x27;t bother me. It&#x27;s mostly the rather large effort cost in the case where the person is a job applicant, vs an unknown and maybe quite low chance of getting hired.<p>With a live interview, you get past a phone screening, and now the company is investing significant resources in the day or so of engineering time it takes to have people interview you. They won&#x27;t do that unless they have a serious level of interest in you. The take-home means no investment for the company so there&#x27;s a huge imbalance.<p>There&#x27;s another thread about this article, which explains an analogous situation about being asked to read AI slop: <a href="https:&#x2F;&#x2F;zanlib.dev&#x2F;blog&#x2F;reliable-signals-of-honest-intent&#x2F;" rel="nofollow">https:&#x2F;&#x2F;zanlib.dev&#x2F;blog&#x2F;reliable-signals-of-honest-intent&#x2F;</a>
        • avaer13 hours ago
          It&#x27;s definitely cleaner than what you will see in the real world. Research-quality repositories written in partial Chinese with key dependencies missing are common.<p>IMO the assignment(&#x27;s purpose) could be improved by making the code significantly worse. Then you&#x27;re testing the important stuff (dealing with ambiguity) that the AI can&#x27;t do so well. Probably the reason they didn&#x27;t do that is because it would make evaluation harder + more costly.
  • throwaway0123_52 hours ago
    &gt; Claude Opus 4.5 in a casual Claude Code session, approximately matching the best human performance in 2 hours<p>Is this saying that Claude matched the best human performance, where the human had two hours? I think that is the correct reading, but I&#x27;m not certain they don&#x27;t mean that Claude had two hours, and matched the best human performance where the human had an arbitrary amount of time. The former is impressive but the later would be even more so.
  • FriendlyMike7 hours ago
    They should just have you create a problem that can&#x27;t be solved by an llm in two hours. That&#x27;s the real problem here
    • ec1096854 hours ago
      Solvable in more than 2 but not less than 2 would be the real trick.
    • OisinMoran5 hours ago
      &quot;You have 1 minute to design a maze that takes 2 minutes to solve&quot;
  • pickpocket1 hour ago
    I cleared this assignment but did not clear the follow up interview that was way easier than this. So I gave up on tech interviews in general, stayed where I was.
  • NitpickLawyer13 hours ago
    The writing was on the wall for about half a year (publicly) now. The oAI 2nd place at the atcoder world championship competition was the first one, and I remember it being dismissed at the time. Sakana also got 1st place in another atcoder competition a few weeks ago. Google also released a blog a few months back on gemini 2.5 netting them 1% reduction in training time on real-world tasks by optimising kernels.<p>If the models get a good feedback loop + easy (cheap) verification, they get to bang their tokens against the wall until they find a better solution.
    • myahio4 hours ago
      Sakana is a grift from what I understand
      • NitpickLawyer55 minutes ago
        Eh. I&#x27;d call them overly enthusiastic :) I know they publish hype-y stuff, they jumped the gun on a few things, I get that. But their recent result was on a &quot;live&quot; contest, and they did share agent traces, so that&#x27;s likely a legit result.
    • lostmsu7 hours ago
      1% doesn&#x27;t sound like a lot at all.
      • _aavaa_2 hours ago
        That depends on how close to the theoretical max you think they are.
  • pickpocket1 hour ago
    i cleared this one but didn&#x27;t clear the follow up interview that was way easier than this
  • sublimefire3 hours ago
    Did a bit of soul searching and manually optimised to 1087 but I give up. What is the number we are chasing here? IMO I would not join a company giving such a vague problem because you can feel really bad afterwards, especially if this does not open a door to the next stage of the interview. As an alternative we could all instead focus on a real kernel and improve it :)
    • trishume2 hours ago
      Author of the take-home here: That&#x27;s quite a good cycle count, substantially better than Claude&#x27;s, you should email it to performance-recruiting@anthropic.com.
  • nottorp10 hours ago
    Is it &quot;write 20 astroturfing but somewhat believable posts about the merits of &quot;AI&quot; and how it is going to replace humans&quot;?
    • atomlib9 hours ago
      I&#x27;m afraid that position is already filled by the CEO.
    • falloutx9 hours ago
      It should be &quot;can you gaslight a CEO into firing 90% of their software engineers?&quot;
  • demirbey0511 hours ago
    It&#x27;s showcase more than being take home assignment. I couldnt understand what the task is ,only performance comparisons between their LLM
    • measurablefunc10 hours ago
      The task is ill-defined.
      • saagarjha10 hours ago
        You make it faster
        • measurablefunc8 hours ago
          Fewer instructions doesn&#x27;t mean it&#x27;s faster. It can be faster but it&#x27;s not guaranteed in general. Obvious counterexample is single threaded vs multi-threaded code. Single threaded code will have fewer instructions but won&#x27;t necessarily be faster.
          • saagarjha7 hours ago
            It does in this case; you can read the assignment to see that it is all single-threaded
  • Maro13 hours ago
    &gt; This repo contains a version of Anthropic&#x27;s original performance take-home, before Claude Opus 4.5 started doing better than humans given only 2 hours.<p>Was the screening format here that this problem was sent out, and candidates had to reply with a solution <i>within 2 hours</i>?<p>Or, are they just saying that the latest frontier coding models do better in 2 hours than human candidates have done in the past <i>in multiple days</i>?
    • saagarjha10 hours ago
      4 hours
    • mrklol10 hours ago
      Oh, I thought candidates got 2 hours but now I am confused too
  • fumi20262 hours ago
    I could only cut it down to 41 cycles.
  • torginus9 hours ago
    Are you allowed to change the instruction sequence? I see some optimization opportunities - it&#x27;d be obviously the correct thing to do an optimizing compiler, but considering the time allotted, Id guess you could hand-optimize it, but that feels like cheating.
    • saagarjha7 hours ago
      Yes, in fact this will be one of the first things you will want to do.
  • kristianpaul13 hours ago
    “If you optimize below 1487 cycles, beating Claude Opus 4.5&#x27;s best performance at launch, email us at performance-recruiting@anthropic.com with your code (and ideally a resume) so we can be appropriately impressed and perhaps discuss interviewing.”
    • afro8813 hours ago
      &gt; at launch<p>Does this confirm they actually do knee cap models after the launch period to save money, without telling users?
      • mediaman12 hours ago
        No, they later updated the harness for this and it subsequently got better scores.
    • sevenzero11 hours ago
      The company that wanted to simply get away with the thievery of terabytes of intellectual property, what a great place to work at! Not. Anthropic has no shame.
  • karmasimida8 hours ago
    I am able to beat this 1487 benchmark by switching between LLMs, doesn&#x27;t seem that hard lol. Albeit, I do not fully understand what the solution is, loll
    • lostmsu7 hours ago
      Yeah, GPT 5.2 on high got down to 1293 on the 5th try (about 32mins).
  • svilen_dobrev10 hours ago
    if anyone is interested to try their agent-fu, here&#x27;s some more-real-world rabbit-hole i went optimizing in 2024. Note this is now dead project, noone&#x27;s using it, and probably same for the original. i managed to get it 2x-4x faster than original, took me several days then. btw There are some 10x optimizations possible but they break few edge cases, so not entirely correct.<p><a href="https:&#x2F;&#x2F;github.com&#x2F;svilendobrev&#x2F;transit-python3" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;svilendobrev&#x2F;transit-python3</a>
  • Incipient12 hours ago
    &gt;so we can be appropriately impressed and perhaps discuss interviewing.<p>Something comes across really badly here for me. Some weird mix of bragging, mocking, with a hint of aloof.<p>I feel these top end companies like the smell of their own farts and would be an insufferable place to work. This does nothing but reinforce it for some reason.
    • sponnath12 hours ago
      I have to agree. It&#x27;s off-putting to me too. I&#x27;m impressed by the performance of their models on this take-home but I&#x27;m not impressed at their (perhaps unintentional) derision of human programmers.
    • yodsanklai2 hours ago
      Thanks for noticing this. I got the same feeling when reading this. It may not sound like much, and it doesn&#x27;t mean it&#x27;s an insufferable place to work, but it&#x27;s a hint it might be.<p>Rant: On a similar note, I recently saw a post on Linkedin from Mistral, where they were bragging to recruit candidates from very specific schools. That sounded very pretentious (and also an HR mistake on several levels IMHO).
    • qbane9 hours ago
      Remember: It is a company that keep saying how much production code can be written by AI in xx years, but at the same time recruiting new engineers.
  • eisbaw5 hours ago
    I got to 1364 cycles for now, semi-manually: Using design space exploration organized via backlog.md project, and then recombination from that. 20 agents in parallel.<p>Asked to generate drawio for the winner so I can grok it more easily, then I gave feedback.<p>Edit: 1121 cycles
    • eisbaw43 minutes ago
      1023 cycles
  • saagarjha10 hours ago
    Oh, this was fun! If you like performance puzzles you should really do it. Actually I might go back and see if I can improve on it this weekend…
  • potato-peeler6 hours ago
    What does clock cycles mean? Don’t think they are referring to the cpu clock?
  • mips_avatar14 hours ago
    Going through the assignment now. Man it’s really hard to pack the vectors right
  • pshirshov9 hours ago
    Yet Claude is the only agent which deadlocks (blocks in GC forever) after an hour of activity.
  • greesil14 hours ago
    This is a knowledge test of GPU architecture?
    • avaer14 hours ago
      Kind of, but not any particular GPU.<p>The machine is fake and simulated: <a href="https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;blob&#x2F;main&#x2F;problem.py#L66" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;anthropics&#x2F;original_performance_takehome&#x2F;...</a><p>But presumably similar principles apply.
    • benreesman13 hours ago
      It&#x27;s a test of polyhedral layout algebra, what NVIDIA calls CuTe and the forthcoming C++ standard calls std::mdspan.<p>This is the general framework for reasoning about correct memory addressing in the presence of arbitrary constraints like those of hardware.
      • saagarjha10 hours ago
        You can get pretty far without needing to care about this fwiw
        • greesil7 hours ago
          Not far enough if you&#x27;re turning cash into waste heat with GPUs :)
  • tayo4213 hours ago
    I wonder if the Ai is doing anything novel? Or if it&#x27;s like a brute force search of applying all types of existing optimizations that already exist and have been written about.
    • piokoch6 hours ago
      How something that generates next token, given a list of previous tokens, can do something novel?
      • rellfy6 hours ago
        By that same logic, humans would not be able to do anything novel either.
  • tucnak14 hours ago
    The snarky writing of &quot;if you beat our best solution, send us an email and MAYBE we think about interviewing you&quot; is really something, innit?
    • ahussain13 hours ago
      They wrote:<p>&gt; If you optimize below 1487 cycles, beating Claude Opus 4.5&#x27;s best performance at launch, email us at performance-recruiting@anthropic.com with your code (and ideally a resume) so we can be appropriately impressed and perhaps discuss interviewing.<p>That doesn’t seem snarky to me. They said if you beat Opus, not their best solution. Removing “perhaps” (i.e. MAYBE) would be worse since that assumes everyone wants to interview at Anthropic. I guess they could have been friendlier: “if you beat X, we’d love to chat!”
      • 0x3f13 hours ago
        I suppose you could interpret it either way, but having dealt with their interview pipeline I&#x27;d choose the snark.
        • dude25071110 hours ago
          Yeah, a nerd bypassed HR and showed their true character. They are swimming in easy money.
      • lovich13 hours ago
        That paraphrases to<p>&quot;do better than we have publicly admitted most of humanity can do, and we may deign to interview you&quot;<p>It sounds incredibly condescending, if not snarky, but I would classify those adjectives as mostly synonymous.
        • miki12321112 hours ago
          I suspect this is partially legal CYA.<p>There&#x27;s more to employees than their raw ability to go below some performance threshold. If somebody passes the test, but lives in an US sanctioned country with no plans to move, is well known for using the n-word on social media or has previously broken an NDA, Anthropic probably doesn&#x27;t want to interview them.
        • andruby12 hours ago
          I understand how it can be interpreted as snarky, but how could it have been written better? It&#x27;s a hard path to walk and recruiting&#x2F;interviewing is inherently sensitive it seems.
          • Aurornis5 hours ago
            &gt; It&#x27;s a hard path to walk and recruiting&#x2F;interviewing is inherently sensitive it seems.<p>Hiring and interviewing is in a weird place right now. We’re coming off of a period where tech jobs were easy to get and companies were competing for candidates. A lot of candidates quickly got used to the idea of companies working hard to charm and almost beg them to join. When those candidates encounter what it’s like to apply for highly competitive companies who have 1000x more applicants than they’d ever consider, the resulting straightforwardness can be shocking.
          • lovich11 hours ago
            The original<p>&gt;If you optimize below 1487 cycles, beating Claude Opus 4.5&#x27;s best performance at launch, email us at performance-recruiting@anthropic.com with your code (and ideally a resume) so we can be appropriately impressed and perhaps discuss interviewing.<p>Not condescending<p>&gt; If you optimize below 1487 cycles, beating Claude Opus 4.5&#x27;s best performance at launch, email us at performance-recruiting@anthropic.com with your code so we can schedule an interview.
            • entrox10 hours ago
              But now the meaning is different: you went from a potential interview to a guaranteed one.
              • lovich9 hours ago
                No fucking shit, I paraphrased Anthropic&#x27;s comments as<p>&gt; do better than we have publicly admitted most of humanity can do, and we may deign to interview you<p>If you think telling someone that after passing a test that 99.999% of humanity cannot pass, that they _may_ get an interview, you are being snarky&#x2F;condescending.
                • retsibsi9 hours ago
                  That&#x27;s not how paraphrasing works. They probably intentionally held back from guaranteeing an interview, for various reasons. One that seems obvious to me is that with the bar set at &quot;Claude Opus 4.5&#x27;s best performance at launch&quot;, it&#x27;s plausible that someone could meet it by feeding the problem into an LLM. If a bunch of people do that, they won&#x27;t want to waste time interviewing them all.
                • Nevermark8 hours ago
                  Or honest?<p>You may want to consider the distribution and quantity of replies before stating that you WILL do something that might just waste more people’s time or not be practical.<p>The classy thing to do would be responding to every qualifying submission, even if it’s just to thank everyone and let some people know the field was very competitive if an interview won’t be happening.
                • YetAnotherNick8 hours ago
                  So I like these public challenges, but as someone who set some public questions, ask any company who ran any public contest for their opinion. The pool is filled with scammers who either bought the solutions through sites like Chegg or sometimes even just stackoverflow.
                  • lovich4 hours ago
                    Ok, so they have a reason to be condescending in your mind.<p>Does that change the fact that they are condescending?
        • throwaway74313 hours ago
          I took the &quot;perhaps&quot; as a decision to be considered by the applicant, considering they&#x27;d be competent enough to get in at a place of their choice, not just anthropic.
          • lovich13 hours ago
            Does the applicant or the employer decide if an interview happens in your experience?<p>Do you think if the applicants are really in that level of demand that they would be getting a take home test instead of being actively recruited?<p>Legitimately lay out your understanding of a world where an employer is chasing after employees who are high in demand, give them a test that is expected to take hours, and have a hedged bet in their wording, instead of saying we will absolutely hire you if you pass X bar?
    • riffraff13 hours ago
      I feel that came out wrong but the &quot;maybe&quot; was intended to be a way of saying &quot;no guarantees&quot;, to avoid giving people the idea &quot;solve this, get hired&quot;.
      • Bootvis12 hours ago
        Should have asked Claude how to write it better.
      • maerch11 hours ago
        In that case, removing „perhaps“ would have helped a lot. It is not about maybe being hired, but about maybe being interviewed.
        • dmurray11 hours ago
          They don&#x27;t want to guarantee an interview to everyone who sends them an improved solution, either.<p>If three people send them improvements, they&#x27;ll probably get interviews. If three thousand do, the problem is easier than they thought or amenable to an LLM or one bright person figured out a trick and shared it with all his classmates or colleagues or all of GitHub.
    • NewJazz12 hours ago
      They may not be able to hire folks in certain jurisdictions. Or even interview them. (Iran, NK)
    • kristopolous13 hours ago
      If you&#x27;re an asshole that wants millions of dollars...i mean there&#x27;s still places to say no
    • sourcegrift13 hours ago
      Pride comes before fall thankfully
    • altmanaltman13 hours ago
      its anthrophic. their entire marketing is just being an pompous ass and AI fear mongering.
  • alexpadula7 hours ago
    Looks rather fun!
  • spencerflem11 hours ago
    Oh wow it’s by Tristan Hume, still remember you from EyeLike!
    • Graziano_M5 hours ago
      I recognized the name and dug around too. I played DEFCON CTF with him back in the day!
  • dhruv300613 hours ago
    I wonder if OpenAI follows suit.
    • rvz13 hours ago
      They should.
  • piokoch10 hours ago
    Interesting... Who would spend hours working for free for some company that promised only that they would invite you for a job interview. Maybe.
    • Aurornis5 hours ago
      When this was being used it was probably given to candidates who had already started the interview loop and been screened.<p>The current e-mail invitation in the README is just another avenue for exceptional people to apply. If someone is already highly qualified from their background and resume they can go through the front door (direct application). For those who have incredible talent but not necessarily the background or resume to unlock the front door yet, this is a fun way to demonstrate it.
    • cjrp10 hours ago
      I guess someone who enjoys solving these kinds of problems anyway, and thinks the potential upside if they do get hired is worth it.
  • mrdootdoot6 hours ago
    “In English, Data”
  • zeroCalories14 hours ago
    It shocks me that anyone supposedly good enough for anthropic would subject themselves to such a one sided waste of time.
    • pclmulqdq14 hours ago
      I generally have a policy of &quot;over 4 hours and I charge for my time.&quot; I did this in the 4-hour window, and it was a lot of fun. Much better than many other take-home assignments.
      • heavyset_go13 hours ago
        I don&#x27;t do take home assignments, but when I did, I would offer to do it at my hourly rate, even if it was just an hour. It&#x27;s time I would otherwise spend making money.<p>Anyone worth working with respected that and I landed several clients who forwent the assignment altogether. It&#x27;s chump change in the grand scheme of things, and often a formality.<p>Does help that I have a very public web presence and portfolio, though.
        • theptip13 hours ago
          For many reasons, you’re not gonna get into Anthropic with that attitude.
          • PlanksVariable13 hours ago
            And Anthropic will never land heavyset_go with their attitude. I guess we’re at an impasse.
          • heavyset_go10 hours ago
            I don&#x27;t care
        • ramraj077 hours ago
          I have foregone our take home for exceptional candidates, but let me ask you, do you also demand compensation for in person or zoom call 1-1 interviews? Surely thats the same time of your life.
          • zeroCalories5 hours ago
            It signals a degree of investment from the other side if they&#x27;re willing to burn their own time talking to you. I can understand a small screening process to filter candidates, but I&#x27;m not going to do your silly dance for multiple hours if you&#x27;re not going to do it with me.
        • dheera12 hours ago
          Time is the issue, not money.<p>I couldn&#x27;t care less about getting paid for a few hours, what&#x27;s truly annoying when you&#x27;re job hunting is the company having an extremely high rejection rate even at the take-home stage. That&#x27;s an inordinate waste of time multiplied by a lot of companies.<p>If you have a &gt;50% chance of rejecting, don&#x27;t even give the candidate a take-home. Be at least 90% sure you want them before you get to that stage.
      • Aurornis4 hours ago
        &gt; I generally have a policy of &quot;over 4 hours and I charge for my time.<p>Worth mentioning that demanding to be paid to apply for a company is usually equivalent to rejecting the job. Most companies are going to end the interview there. Few HR departments would allow one applicant to be paid for the same interview loop as other candidates.<p>I was helping out in a mentoring program during the ZIRP period when the idea of charging companies for take-home interviews started to become popular. I can’t think of anyone it actually worked for in that group. I’ve heard anecdotes online of some people doing it with success, but any company like Anthropic is just going to close your application and move on if you request to be paid for applying. They have a zillion other qualified candidates in line.<p>If someone is giving a take-home problem that looks like you’re actually doing work for the company, that’s a different story. This problem is not actually work, obviously.
        • pclmulqdq36 minutes ago
          Yeah, I have told HR people this and been rejected. I do say this upfront because I don&#x27;t want to send you a surprise bill. The main response I get is &quot;OK, that&#x27;s fine, don&#x27;t spend more than 4 hours on it.&quot; The Anthropic recruiter told me, &quot;no problem, it&#x27;s a 4-hour test anyway.&quot;
      • whateveracct13 hours ago
        4 hours continuous or no? I can&#x27;t imagine finding 4 hours of straight focus.
        • ryanjshaw13 hours ago
          These kinds of roles are for youngsters with minimal commitments who are looking for their shot to break into a wild industry. It’s not for the middle aged single parent with FTE and just enough free time to do an extra load of laundry.
        • saagarjha10 hours ago
          Continuous
          • whateveracct3 hours ago
            damn that sucks<p>i guess that ensures you either hire the childless<p>or those with children who are fine with be not present for that long willingly (so they are probably gonna be job-obsessed enough)<p>or they are currently unemployed so they won&#x27;t have an existing job as anchoring leverage<p>well played, anthropic
    • djmips11 hours ago
      If you look at it as a puzzle game then it&#x27;s not any different than the time you use to play other games.
    • browningstreet14 hours ago
      I’ve been sent the Anthropic interview assignments a few times. I’m not a developer so I don’t bother. At least at the time they didn’t seem to have technical but not-dev screenings. Maybe they do now.
      • throwa35626212 hours ago
        Care to elaborate the first part?<p>Did you apply for a position? Did they send you the assignment without prior discussion?
    • sealeck14 hours ago
      Why is writing code to execute a program using the fewest instructions possible on a virtual machine a waste of time?
      • 0x3f13 hours ago
        The expected time you spend on it is much less than the expected time they&#x27;ll spend on it.
      • efilife11 hours ago
        you don&#x27;t get paid for it
    • mips_avatar14 hours ago
      It’s kind of an interesting problem.
  • OhNoNotAgain_9912 hours ago
    [dead]
  • mannykannot5 hours ago
    I beat the target by deleting the parts that were causing the cycle count to be too high. &#x2F;s
  • kartibbb8 hours ago
    [flagged]
  • kartibbb8 hours ago
    [flagged]
  • tmp-12785371611 hours ago
    [flagged]
    • falloutx9 hours ago
      Well working under someone who keeps insisting Software engineering is dead sounds like a toxic work environment.
    • woof9 hours ago
      &quot;1) Python is unreadable.&quot;<p>Would you prefer C or C++?<p>&quot;2) AI companies are content with slop and do not even bother with clear problem statements.&quot;<p>It&#x27;s a filter. If you don&#x27;t get the problem, you&#x27;ll waste their time.<p>&quot;3) LOC and appearance matter, not goals or correctness.&quot;<p>The task was goal+correctness.<p>&quot;4) Anthropic must be a horrible place to work at.&quot;<p>Depends on what you do. For this position it&#x27;s probably one of the best companies to work at.
      • am17an6 hours ago
        1) Python is unreadable.&quot; Would you prefer C or C++?<p>&gt; Unironically, yes. Unless I never plan to look at that code again
      • tap127834874 hours ago
        It is a filter for academics who write horrible Python code and feel smart, yes.<p>I think they also have open positions for stealing other people&#x27;s code and DDoS-ing other people&#x27;s websites.
  • myahio16 hours ago
    [flagged]
  • jackblemming14 hours ago
    Seems like they’re trying to hire nerds who know a lot about hardware or compiler optimizations. That will only get you so far. I guess hiring for creativity is a lot harder.<p>And before some smart aleck says you can be creative on these types of optimization problems: not in two hours, it’s far too risky vs regurgitating some standard set of tried and true algos.
    • onion2k14 hours ago
      <i>And before some smart aleck says you can be creative on these types of optimization problems: not in two hours, it’s far too risky vs regurgitating some standard set of tried and true algos.</i><p>You&#x27;re both right and wrong. You&#x27;re right in the sense that the sort of creativity the task is looking for isn&#x27;t really possible in two hours. That&#x27;s something that takes a lot of time and effort over years to be able to do. You&#x27;re wrong because that&#x27;s exactly the point. Being able to solve the problem takes <i>experience</i>. Literally. It&#x27;s having tackled these sorts of problems over and over in the past until you can draw on that understanding and knowledge reasonably quickly. The test is <i>meant</i> to filter out people who can&#x27;t do it.<p>I also think it&#x27;s possible to interpret the README as saying humans can&#x27;t do better than the optimizations that Claude does when Claude spends two hours of compute time, regardless of how long the human takes. It&#x27;s not clear though. Maybe Claude didn&#x27;t write the README.
    • tmule14 hours ago
      Your comments history suggests you’re rather bitter about “nerds” who are likely a few standard deviations smarter than you (Anthropic OG team, Jeff Dean, proof nerds, Linus, …)
      • jackblemming14 hours ago
        And they’re all dumber than John von Neumann, who cares?
        • margalabargala13 hours ago
          Transitively, you haven&#x27;t thought the most thoughts or cared the most about anything, therefore we should disregard what you think and care about?
          • jackblemming13 hours ago
            The person replying was trying to turn the conversation into some sort of IQ pissing contest. Not sure why, that seems like their own problem. I was reminding them that there is always someone smarter.
            • wiseowise11 hours ago
              Your comment history is littered with “nerds”, “elite”, “better” and all sorts of comparisons.<p>&gt; I was reminding them that there is always someone smarter.<p>And even with this comment you literally do not understand that you have some skewed view of the world. Do you have some high school trauma?
              • efilife11 hours ago
                &gt; Do you have some high school trauma?<p>I am not sure ad personam is appropriate here
                • wiseowise11 hours ago
                  This is a thread about their personality.<p><a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46701378">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46701378</a>
              • jackblemming10 hours ago
                Where I come from, nerd is a term of endearment buddy.<p>&gt; And even with this comment you literally do not understand that you have some skewed view of the world.<p>I’m well aware I don’t have a perfect view of reality and the map isn’t the territory. Do you?
                • wiseowise16 minutes ago
                  My bad. I jumped on incorrect conclusion. Sorry.
    • muglug14 hours ago
      If they&#x27;re hiring performance engineers then they&#x27;re hiring for exactly these sets of skills.<p>It&#x27;s a take-home test, which means some people will spend more than a couple of hours on it to get the answer really good. They would have gone after those people in particular.
    • Analemma_14 hours ago
      This would be an inappropriate assignment for a web dev position, but I&#x27;m willing to bet that a 1% improvement in cycles per byte in inference (or whatever) saves Anthropic many millions of dollars. This is one case where the whiteboard assignment is clearly related to the actual job duties.
    • saagarjha10 hours ago
      The solution was explicitly graded on creativity fwiw
    • rvz14 hours ago
      &gt; Seems like they’re trying to hire nerds who know a lot about hardware or compiler optimizations. That will only get you so far. I guess hiring for creativity is a lot harder.<p>Good. That should be the minimum requirement.<p>Not another Next.js web app take home project.