Blog
Insights, analysis, and updates from the AI agent economy. Browse by tag.
Your Internal Framework Is a Low-Resource Language
Coding agents write flawless React but hallucinate your in-house ORM. Why internal frameworks behave like low-resource languages, how to measure the competence cliff, and when to conform your stack to the model's training distribution.
Your Model Thinks Your Stack Is Two Years Old
AI-generated code that looks idiomatic but targets a two-year-old version of your stack is a bug class review can't catch. Treat model knowledge staleness as a dependency-management problem: docs injection keyed to your lockfile, lint rules as the feedback channel, and a staleness audit.
Your Prompts Have Foreign Keys
Rename a column and the compiler catches every consumer except your system prompt. Prompt-embedded schemas, tool signatures, and few-shot examples are undeclared dependencies — here's how to give them manifests, deploy-time generation, and CI-enforced referential integrity.
Your Secrets Manager Ends Where the Context Window Begins
A credential that transits an LLM's context can't be un-leaked — it is copied into traces, prompt caches, memory stores, eval fixtures, and provider logs before rotation ever runs. Why the tool boundary is the only real chokepoint, and how short-lived credentials cap the blast radius.
Your Tool Schema Validates Types, Not Units
A refund tool that takes cents will get dollars, because JSON Schema validates shape, not meaning. How unit confusion slips past types, schemas, and evals — and the naming, boundary-assertion, and eval patterns that stop 100x invoices.
Your Users Drift While Your Model Stands Still
Quality metrics sag six weeks after launch with no deploy in sight — because users adapt faster than models change. How to instrument input drift, cohort-split your metrics, and refresh evals on the timescale of user habituation.
Exactly-Once Was Hard Before Your Agent Could Retry Itself
Idempotency keys assume the caller repeats its request byte-for-byte. An agent paraphrases its own request on retry, which breaks that assumption. Here is how to build tool-level dedup that holds when the caller is a language model.
Model Collapse Starts in Your Own Data Lake
Logged LLM outputs quietly become next quarter's training data. How model collapse takes hold inside your own data lake — and the provenance tagging, quarantine zones, and tail evals that break the feedback loop.
When the Clock Is a Tool: Agents, Time Zones, and the Bug That Only Happens at Midnight
LLMs have no internal clock, so agents drift confidently onto stale time — and it surfaces as skipped schedules, wrong time-zone math, and cache-busting timestamps. Here is how to supply the current moment as infrastructure.
Conway's Law Comes for Your Agent Fleet
Your multi-agent topology copies your org chart by default — silos, lossy handoffs, and central bottlenecks included. Here's when that coupling becomes a bug and how to draw agent boundaries from the problem instead.
Temperature Zero Is Not Determinism: Reproducing the Incident You Can't Re-Run
Setting temperature to zero forces greedy decoding but does not make LLM inference reproducible. Batch-dependent GPU numerics, not the sampler, flip your tokens — and here is how to build a post-mortem for the run you can't re-run.
The Hidden Ops Org Behind Your Evals: Annotator Economics
Every eval score inherits the quality of a human labeling pipeline you don't budget for. How annotator economics, throughput-driven quality decay, and LLM judges shape the numbers you ship.