25 POSTS
A skill that reviews a workspace against someone else's methodology is only as honest as its sources. Notes on building one from quoted records, freezing the answer key before writing the skill, and what a 4 of 9 first run taught me.
The claim that the harness matters more than the model is repeated everywhere. I went through what my sources actually show, what they only assert, and why a same-prompt head-to-head cannot carry the conclusion people draw from it.
When agents write the code, the scarce thing is knowing whether it is right. Practitioners handle that with deterministic gates, review depth scaled to blast radius, and a reviewer that did not write the code. Most of it is judgment, not measurement.
A coding harness can gate on tests. A harness for slides or design cannot, so it has to turn taste into reference sets, rubrics, constrained components and a human pick. Gamma's own account is the worked example, with the founder's claims marked as claims.
Agents make code cheap to write, so the scarce thing is how much complexity a codebase can carry. Lee Robinson's argument for spending abstractions like a budget, and where Ousterhout-style design fundamentals agree with it.
Before picking a model or a framework, sort each decision in the workflow: a person, a plain rule, classical ML, or a generative model. Every wrong assignment has its own cost, and the costs run in both directions.
AI-assisted CAN bus reverse engineering works when the model sits inside a pipeline of deterministic decoding and validation. It fails when the model is asked to guess signals from raw frames.
There are three ways to give an agent a tool. The choice mostly comes down to who else needs the tool and what it costs in context, not which one is best.
A skill that holds up is built in a fixed order: do the task by hand, write down what you did, turn it into a skill, then test it against cases. Most broken skills skipped a step.
What makes a coding agent reliable is mostly the layer around the model: the context it is handed, its tools, the checks on its output and the code that orchestrates sessions. That layer is ordinary engineering.
An agent loop is a trigger, an action and a stop condition. The stop condition is the only hard part, because done has to be something a machine can check.
When I cannot judge a plan myself, the model that wrote it should not be its only reviewer. A model from a different vendor can critique it in a bounded loop, and any factual claim the reviewer makes still has to be checked.
Running a coding agent fully offline is practical now. The binding constraint is GPU memory, and the honest question is what quality you give up at each memory tier.
The CoALA taxonomy names four kinds of agent memory: working, semantic, procedural and episodic. In practice each one is a plain file, and knowing which kind a fact is tells me which file it belongs in.
A skill library gets worse past a certain size, because everything in it competes for the model's attention. Deleting is a real optimization, and the test for keeping a skill is whether the agent does better with it than without it.
Compaction keeps a general summary that the tool chose. A handoff document keeps what I decided the next session needs. They solve different problems, and the handoff is the one worth writing by hand.
Generative UI lets the agent choose layout and components instead of filling a fixed template. Notes on Cole Medin's three tiers: static, declarative and open-ended.
Greg Isenberg and Ras Mic on how context windows actually work, why most agent setups waste them, and why skills do more than a long CLAUDE.md.
Ras Mic walks through documentation generated from Claude Code's leaked source: the agentic loop, how context is assembled, how tools run and how permissions work.
Nate B Jones on why agent success stories do not transfer: output scales far faster than an organization's ability to review it, and the foundational work still decides the result.
Cole Medin's architecture for a personal second brain on Claude Code and Obsidian, and his security argument for building your own instead of deploying an off-the-shelf agent.
Andrej Karpathy on the No Priors podcast: code agents and AutoResearch loops mark a phase change in how software and AI research get done, which he calls the loopy era.
Released an open-source template that uses folder structure as AI agent architecture -- model-agnostic, zero dependencies, and grounded in a published research paper.
Built a fully local AI resume toolkit with dual ATS + HR scoring, automated job discovery, and DOCX generation.
Agents keep getting better at executing tasks while still lacking persistent organizational context. Notes on Nate B Jones's argument that the fix is context stewardship, not more code.