Your point that effort controls and agent teams matter more than benchmark scores is exactly right. The /insights command and workflow infrastructure upgrades are what actually enable production deployment at scale, not incremental MMLU improvements.
I was doing some experiments with Agents Swarm and /fast - there is a reason why it's in PREVIEW. Plys it is so TOKEN HUNGRY. But - I can see use cases. Like - I have a big codebase and I am looking for a bug - I can run 4 agents on different parts to find it faster.
Also good with research and custom tool calling.
PS. /insights were very good for me. I improved so many things because of that!
so. much. gold. appreciate you bumping this feature. just glancing through right now, that report is superb - insights plus prompt suggestions for improvement. !!! anthropic just keeps on crushing it.
You literally said it right man, "those insights command literally roasted my way of working hahaha". I am eager to try the Agents team, but I think I am still not got into that use case as you have defined that when not to use the Agents team. I think pretty much current building tasks are stuck over there, but I hope I soon get to try it.
Your point that effort controls and agent teams matter more than benchmark scores is exactly right. The /insights command and workflow infrastructure upgrades are what actually enable production deployment at scale, not incremental MMLU improvements.
I looked at what features drive competitive advantage: https://thoughts.jock.pl/p/ai-agent-landscape-feb-2026-data
Actually, it would be cool to see those features being used in real cases! Thank you for reading
I was doing some experiments with Agents Swarm and /fast - there is a reason why it's in PREVIEW. Plys it is so TOKEN HUNGRY. But - I can see use cases. Like - I have a big codebase and I am looking for a bug - I can run 4 agents on different parts to find it faster.
Also good with research and custom tool calling.
PS. /insights were very good for me. I improved so many things because of that!
I’m releasing a post on /insights command in a couple of days, I feel like it is a very useful feature!
And yes, I feel like Opus 4.6 is hungry just in general, but that’s what I heard about “fast” feature and agent swarms - they eat a lot of tokens.
Great breakdown Ilia! Especially as you include the when-not-to-use-it section, that is what people really need to understand.
Appreciate the kind words. Thank you!
great stuff. anthropic killed it.
Agreed. Thank you for reading!
This is a very useful breakdown, thank you!
Of course you’re welcome!
Ooh - the insights command, didn’t know about that one. Going to fire up a new terminal window right now….
Go for it, it’s pretty cool! I’ll be talking about it in my next post too
so. much. gold. appreciate you bumping this feature. just glancing through right now, that report is superb - insights plus prompt suggestions for improvement. !!! anthropic just keeps on crushing it.
Absolutely! And there’s a lot of gold in that report. I’m looking forward to talking about it
You literally said it right man, "those insights command literally roasted my way of working hahaha". I am eager to try the Agents team, but I think I am still not got into that use case as you have defined that when not to use the Agents team. I think pretty much current building tasks are stuck over there, but I hope I soon get to try it.
Thank you for reading! I’m also looking for a use case to try out Agents team, but very much looking forward to it too.