Run AI agents in production without burning money or trust.
I'm Alexey. I write about the unglamorous side of agents: cost control, reliability, and tooling — including on-chain agents. Every post ships a small, runnable tool you can paste and run with no API key, plus the real output.
Recent posts
-
Your Guardrail Test Still Passes With the Guardrail Deleted
Test whether an AI agent guardrail is still working: rerun the corpus with the gate deleted. Each bad input lands LOAD_BEARING, SHADOWED or MISSED.
-
Unbounded LLM Call Sites: A Budget Is Not a Ceiling
Unbounded LLM call sites: a per-call budget is not a ceiling. An offline stdlib ast gate prints BOUNDED, UNBOUNDED or CANNOT ANALYZE for each call site.
-
Rate vs Volume Attribution: $40.80 Disputed, or $0.00
Rate vs volume attribution: a 30-day export shows $40.80 of a $54.50 delta disputed at two snapshots, $0.00 at daily width. Offline gate, real exit codes.
-
Lost Update: When Two AI Agents Edit One File, One Silently Wins
Two AI agents edit one resource. Both ACK, one vanishes. I ran 5 agents on one file: 4 contributions gone. A pre-write compare-and-set gate, measured.
Building a tool for AI-agent developers?
This is a small, niche blog for developers building and operating AI agents. If your product serves them, you can sponsor a post or a placement — always disclosed, always something I can actually run. The tools here stay free to read.