标签：#c

A lot of AI memory systems start with the same temptation: "Just save the useful thing." That sounds harmless until the knowledge base becomes a junk drawer. Half the notes are too specific, a few are duplicates, some are obsolete, and nobody knows which ones the agent should trust. In ai-assistant-dot-files , the memory system is deliberately slower. It uses a promotion lifecycle: Capture -> Candidate -> Audit -> Approve -> Index -> Retrieve -> Expire That lifecycle is documented in docs/runbooks/memory-engineering.md , and the important word is not "capture." It is "candidate." Nothing writes directly to memory The framework has a durable memory layer: Knowledge Items in shared/knowledge/ , ADRs in docs/adrs/ , the domain dictionary, team topology, a feature archive, and a registry at shared/memory-registry.json . But a lesson from a delivery does not jump straight into shared/knowledge/ . It first becomes a Candidate Record. That record has required fields: Source Type Evidence Tags Expiration condition Then memory-engineer audits it: Is it reusable? Is it already covered? Is it too speculative? Does it belong as a Knowledge Item, or should it become a rule change, prompt edit, or ADR instead? Only after that does a human approve the destination. The design is intentionally similar to code review. Durable memory changes future behavior, so they deserve a paper trail. Rejection is a feature One of my favorite parts of the memory runbook is that it has explicit rejection rules. Do not promote a memory when it is: a one-off already covered too speculative That makes "zero candidates promoted this cycle" a healthy result, not a failure. This is where memory engineering starts to look less like note-taking and more like gardening. The point is not to preserve every leaf. The point is to keep the soil useful. Expiration matters The lifecycle also includes expiration. A Knowledge Item can become stale when the underlying code, agent, or pattern changes. It can be supers

React Doctor es un linter de cero instalación que escanea un codebase de React buscando problemas de correctitud, seguridad, accesibilidad, rendimiento y mantenibilidad, luego los rankea y te entrega una lista de arreglos con forma de agente. Este post es un reporte de campo: lo corrí sobre una SPA real de React 18 + Vite + TypeScript (~50 rutas, TanStack Query, react-hook-form) y separé lo que de verdad me atrapó. El resultado honesto: 1,249 hallazgos, cinco que valía la pena arreglar, incluyendo dos bugs de seguridad reales. Los otros 1,244 fueron una mezcla de ruido, decisiones de criterio, y falsos positivos. TL;DR Categoría Reportados Vale la pena actuar Por qué la brecha Seguridad 18 warnings 2 13 eran sinks saneados en el servidor (falsos positivos); 1 mina de código muerto + 1 XSS real fueron el oro Bugs 24 errores, 487 warnings 2 1 fuga de timer, 1 key/spread; ~27 exhaustive-deps piden criterio humano Accesibilidad 434 warnings, 1 error 1 el 1 error ( aria-selected faltante) era real; los 434 son ruido de <Label> / <button type> Rendimiento 110 warnings 0 nada caliente; candidatos pero sin impacto medido Mantenibilidad 176 warnings 0 opiniones de estilo tipo "componente grande" Total 1,249 5 señal-a-ruido ≈ 1 en 250 El puntaje que imprimió: 39 / 100 . Dos cosas que ese encuadre esconde: los cinco que sacó a la superficie eran de alto valor (una fuga de token en localStorage y un sink de XSS almacenado), y el conteo no es determinista : una segunda corrida del mismo commit reportó 1,254, no 1,249. El veredicto en una línea: excelente generador de hipótesis, pésima compuerta. Úsalo para encontrar candidatos, verifica cada uno contra el código, y nunca conectes el conteo al CI. Qué es React Doctor Un solo binario que corres por npx / pnpm dlx , sin agregar dependencia a tu proyecto, sin archivo de configuración obligatorio. Parsea tu src/ , hace match contra un conjunto de reglas (sus familias: Seguridad, Bugs, Accesibilidad, Rendimiento, Mantenibilidad), e im

We deployed a coding agent that hit 94% on the industry benchmark. It failed in production on the first real edge case because the benchmark measured single-turn success and our actual work was multi-turn refinement. The model could not update its beliefs correctly when new evidence arrived, something no single-turn eval would catch. This is not a hypothetical. I have watched agents shine in demo and disintegrate on the messy input that production actually serves. The gap between what we measure and what ships is real, and it is where reliability lives or dies. The benchmark misses the point FutureBench evaluates agents by asking them to predict events that occurred after their training cutoff. This removes the possibility of correct answers coming from memorized training data rather than genuine reasoning. The design matters because it tests whether an agent can reason, not whether it can recall. BayesBench showed that standard LLM evaluations score only final-turn answers in single-turn format, leaving multi-turn belief updating entirely unexamined. Across seven models, scaling improves latent inference and evidence accumulation but LLMs do not match rational Bayesian updating. In production, your agent runs many turns. The benchmark that stops at turn one is not measuring the thing that actually breaks. KINA identified three systematic flaws in knowledge benchmarks: scaling-driven designs that ignore disciplinary representativeness, flat-payment annotation that permits lazy consensus among annotators, and unaudited ranking instability under bounded test budgets. The top model reached 53.17% on an 899-item benchmark across 261 disciplines. That is not saturation. That is headroom. The demo lied I worked with a team that deployed an agentic document processing system. The demo on ten handpicked cases was flawless. The first week of production, it hit an input format the training data never saw, and the system failed silently. No error was raised. The output looked

A new generation of voice models for natural human-AI interaction, now powering ChatGPT Voice.

Amid public backlash over its smart glasses, Meta announced that it will be updating its glasses with a new feature that will disable the camera when it detects that someone has tampered with or destroyed the glasses' privacy LED light. The update is meant to address modders who have taken actions such as physically drilling […]

Chat with Notion Agents anytime Discussion | Link

Everything worked. I'd wired up useOptimistic on a task list, the checkbox flipped the instant you...

I've been slowly adding some new free tools to Moxie Docs (partly for SEO, partly to illustrate some of our feature sets before any commitment) for some reason this mermaid editor one blew up on Google rankings so I figured I'd share in case people find it useful! We also have ADR, AGENTS.md, LLMs, and a few other free tools.