Two agents, one working tree, one disaster
Parallel agents sharing a working tree corrupt each other silently. Worktrees fix it, and cost 4.3 seconds and 364 MB each — not the 200ms the contract claims.
agents
46 posts
Reading back is not optional
An agent reporting success has usually verified the call returned, not the artifact. Reading its own output back is the check it skips most often.
An agent drove a KMP app to the Play Store. The codegen was the easy half.
Autonomous codegen produces an app that compiles; what makes it one that ships is a gate stack refusing any claim that carries no evidence.
Prompt injection does not go away because the model is local
A denylist of three dangerous commands was bypassed by a fourth. Why injection defence has to be allowlist-shaped, and what on-device quietly takes away.
Where your AI coding bill actually goes: I measured one long session
22 articles and 110 generated assets in one session. What the retries cost, and the one change that cut attempts per asset from roughly 1.8 to 1.1.
Running an AI agent in CI without it spending your budget
An agent in CI cannot be asked a question. Budget caps that actually throw, local reproduction before pushing, and the gate that degraded instead of failing.
Agent memory that is just files
Most questions an agent asks about past work are exact, not semantic. What a directory tree answers well, and the one query shape it cannot answer at all.
What should an AI agent never be allowed to do?
A permissions model derived from incidents rather than imagination: three tiers, human-only bypasses, and capabilities removed from reach entirely.
The most useful thing my agent does is refuse
Refusals are only useful if they name the remedy. Four that fired during one session, what each redirected me to, and the one that fired on nothing at all.
The agent that writes the check that blocks it
When the same agent writes the fix and the gate, the gate agrees by construction. RED before GREEN is what makes a machine-written check mean anything.
An agent that installs the app and looks at the screen
The capture loop that makes an on-device claim checkable: force-stop, install, launch, capture, hash. Why the order is load-bearing and what it cannot see.
Put the control flow in code, not in the prompt
Sequencing belongs in a script and judgement in the model. One routine commit cost 180k tokens as reasoning and seconds as bash. Here is the line.