A daily-updated index of deployment, RAG, agent and LLMOps tooling — plus field notes on how AI actually gets landed inside real companies.
Hybrid code review that puts deterministic rules first and the LLM second — the only ordering that survives an enterprise security review.
Git worktree plumbing for running several coding agents at once without them wrecking each other's working tree.
A curated set of reusable skills for coding agents — the difference between an agent that finishes tasks and one that finishes them your way.
The terminal-native agent that made agentic coding mainstream — and the reference point every other coding agent is measured against.
Runs many coding agents in parallel while still using the subscriptions you already pay for — the cost model is the product.
A routing layer that keeps your coding agent alive when one provider rate-limits you — worth it for the fallback alone.
Strips the tell-tale rhythm of model output from generated text — useful, and a reminder that the tells are now well known.
An explicit, readable list of the twenty-odd patterns that make text read as generated — better as a checklist than as a filter.
The shortest distance from pip install to a working retriever — which is why nearly every RAG tutorial starts here.
Validates model output against a schema instead of hoping — and re-asks when validation fails.
Gives an agent continuity across sessions — the missing piece once users expect it to remember them.
LLM traces as standard OpenTelemetry spans — so they land in the observability stack you already pay for.
Self-hosting beats the API on cost per token only above a utilisation threshold almost nobody reaches. Below it you are paying for idle silicon and engineer attention — and only o…
Fine-tuning encodes behaviour; retrieval supplies facts. Asking which is better is asking the wrong question — ask how often the underlying knowledge changes and whether you must…
Every multi-tenant RAG eventually discovers the same thing: filtering results after retrieval is not isolation. The filter has to constrain the search, not trim its output.
Published benchmark tables compare recall on public datasets, which is not your dataset, your filters, or your query distribution. Four operational questions decide it instead.
The risk is not that the model says something wrong. It is that the agent does something wrong — with credentials, in a system that does not have an undo.
For two years the advice was to write better prompts. That advice stopped paying off somewhere around the point where the model became reliably good at following instructions. Wha…
Chunking strategies, reranking setups, evaluation harnesses and the seven failure modes that kill every RAG demo in week three.