Subagents are the biggest cause of AI failures In most harnesses: 1. Main agent writes a summary of what it knows 2. Subagent starts from that summary only, and makes assumptions 3. Subagent works and sends back a new summary of what it did Same flaw as compaction: all summaries are lossy. This is how your assistant makes emails you never wanted, or purchases what you told it just to get the price of So I tried something new: checkpoint agents The initial agent has complete context of what needs to be done. And that input is already cached Fork the session where it is. Send the fork ahead to do the task and it will do exactly what the main agent would have done; their context is identical What is even better is when the fork finishes, the cache of the main agent is kept warm by the subagent (they share the same prefix) I measured it on a 137k-token session with Opus 5.5: → Main agent does it on its own: 67 s, $0.38, context +8-17k → Checkpoint: 117 s, $0.68, context +2k → Classic subagent: 123 s, $0.57, context +4-5k But here is what I found: avoiding large input is even more important than keeping cache. And most subagent tasks were done better either in series with retrieval, or split up to be parallel over fresh sessions from the start I do not think there is a place for subagent delegation at all. There should either be a super-agent that orchestrates and does no work, or a long running agent that can retrieve from previous sessions, not a worker that completes tasks partially and delegates the rest to other workers Don't design harnesses based on our assumptions of human behavior

Compaction is a mistake. When an agent's context fills up, tools like Codex and Claude Code compact it: 1. The model prefills the entire session again 2. It writes a summary of everything so far 3. A new session starts from that summary I measured this on a real ~400k-token session with Opus 5.5: → 38 s and $1.99 when the prompt cache has expired → 29 s and $0.17 when it hasn't And the summary is lossy. The exact error message, the file path, the one number you needed is often gone. So I do something new for Mainplane Context retrieval works the other way around. Mainplane keeps the full record of every session as a file. When the context fills up, a new session starts right away with one instruction: "You are continuing session X. Retrieve key context from it as you go." The agent searches the old session and copies the exact lines it needs, when it needs them. Same session, same model: 15 s and $0.10 to get back to work. It does not depend on a cache hit, and nothing is paraphrased. This is the bitter lesson applied to agent memory. Don't hand-design what the model should remember. Give it the full record and good tools, and let it decide. Mainplane is open source (MIT). Link in the first comment.

Introducing: Mainplane🛫 A new open-source AI agent I built from scratch in Go Local agents die with your laptop. Cloud agents can't reach your work. Mainplane lives on your server and drives all your computers > Your old computers → always-on AI shed > Shared file server for all computers > Use a Claude/ChatGPT plan on regular usage limits > Browser + computer use on Mac, Windows and Linux One line to install 100% open source, MIT ↓

It's crazy what my autonomous agents have figured out By starting with browser-use + agentmail + agentcard they: - used email + wallet to buy a joltSMS phone number - email + phone number gets you a apple account - apple account gets you an X account - buying X premium gets you developer API - email + sms also gets you a github - with apple, X, and github they can sign up for almost any service Its essentially a whole digital identity, and they found this path all on their own with one prompt. What else can they do?

We talked with hundreds of users, and turned the hardest real user tasks into a browser-use benchmark We invested heavily into atomic, verified, unambiguous rubrics for LLM judges. This is the best browser agent benchmark ever created 🧵

Made some improvements to bcode + local models Biggest winner is Laguna S2.1 which was compromised by provider issues, it gained +23% score since last post, now is very competitive

Our new cloud agent (live tomorrow) can make posters! Its so much nicer to see information in a styled page than plaintext. It can actually do a lot more than just posters, but more on that later... Try it tomorrow!