> 5 phases and 33 stages from initialization through operation<p>> 14 agents: 11 domain experts, 2 reviewers, and an adaptive composer<p>wow, that's a lot of domain experts, I'm convinced
Not unless you look at my framework (Which I've just hacked together after looking at this) - it has 12! domain experts!
> The current recommended model is Claude Opus 4.8.<p>Staying well behind the frontier, just in case.
It was like that time I nearly dropped out of undergrad, but it was my mistake, I had just forgotten that I was actually a doctorate.
It's a great for tokenmaxxing /s
I'm not sure that it needs to be this complicated. But yes, I am interested in the problem of hard guardrails between multiple agents, and it is not an easy problem to solve.
Sounds good in theory, but are there evaluations of whether or not it actually works?
The answer to that is no. A lot of these are released on vibes and showing that it does what it says it does. For most organizations that care about processes and procedures thats... enough? I've worked on agentic development in enterprise and they just want their way to stamp their old methods on to the new agentic world so it can pass through their release hoops.<p>I've tried to benchmark them the most common public skills including those for ADLC (ie SuperPowers) couple of times here <a href="https://orcabot.com/benchmarks/agent-skills" rel="nofollow">https://orcabot.com/benchmarks/agent-skills</a>. Annoyingly I haven't been able to release one in a while because its been next to impossible to stop the new frontier models from cheating one way or another.
Probably the greatest problem in AI right now is proof / benchmarking<p>I check every benchmark site weekly, compare harnesses and models, and its still pretty much just vibes
My org has something similar.<p>Evals showed that speed and token use improved when it was removed.....<p>The models have improved and internalized some of these principles.<p>They still have value, IMO, because they help improve consistency (at the expense of speed and cost).
I am one of the organisational victims who have been trialling this.<p>My assessment has been just use the native harness.
Well, if it was any good we know for sure that they didn't use it to generate their README.MD with it.