Blog
Topic roundups written from the agentique feed.
-
· prompt-engineering
Why I moved agentique's LLM layer to BAML
Structured output, prompt tests in VSCode, and provider fallbacks: how BAML replaced ad-hoc LLM code in agentique's news pipeline.
-
· security
The security boundary was never where you thought
A CLI privacy toggle that doesn't touch its own upload path, a kill switch inside the process it's meant to stop, and a sandbox that misses one path.
-
· tool-calling
The tool-calling glue is coming from GitHub, not labs
A cost-cutting Claude Code orchestrator, a geospatial SQL skill, and a Claude-Codex review loop, all built by individuals, not a model vendor's product team.
-
· local-ai
Your GPU budget is behind, not the models
A 7MB embedding model running in the browser, a 1-petaflop desktop chip, and a GPT-2 built from scratch in pure C.
-
· model-releases
Open weights stopped being the compromise option
Claude Sonnet 5, GPT-5.6, and an open-weight model with a 1M-token context that runs on any Nvidia GPU.
-
· rag
The RAG problem nobody likes to admit
A 0.6B model doing 100k searches a second at LinkedIn, and why throwing away 68% of your retrieved context is the fix.
-
· agents
Agents got cheaper. Someone wants to break them.
Cheaper multi-step browsing, a 60% token cut for agent workloads, and a CTF built to compromise an agent in a microVM.
-
· coding-assistants
What I'd actually try from this week's coding tools
A wiki that writes itself, a 49% cut in prompts per task, and a code reviewer you can actually self-host.