A local Antares-1B reproduction separates finding a vulnerable file from knowing when to stop reporting it, using paired vulnerable and patched snapshots.
AI · benchmarking · security7 min readThe real stratification in the AI era has nothing to do with how well you code. Your ceiling for safely using AI equals your ability to verify what it produces — a quantity you can actually compute, and one that explains why AI conquered Go a decade ago while still causing accidents in law.
ai · engineering · methodology12 min readAI that writes, runs, and fixes its own code versus AI that hands you a log to read: the output differs by orders of magnitude. The gap does not come from the model. It comes from whether you gave it an exit code it can read by itself.
ai · engineering · methodology10 min readPart two of the Alkaira series. One AI agent claims to run an underground intelligence network. Another believes she controls his informants. The ledger recognizes neither. On why lying should be allowed, and how a world defends the facts.
ai · agent · simulation8 min readA long-running AI society experiment: 16 characters cultivating, scheming, trading, aging, and dying inside the same demonic sect. What I saw, and why the hardest part has nothing to do with intelligence.
ai · agent · simulation11 min readFour AI agents could collaborate on code, but operating the team was still a manual job. So I built one dashboard for deterministic routing, task tracking, and context-window monitoring.
ai · agent · engineering6 min readWhen AI becomes the new Jurchen cavalry, are you the doomed emperor or the one building a new empire?
AI · career · history12 min readFive real options from $249 to $1,399, what each one gives up, and why I ultimately chose something else.
ai · homelab · hardware11 min readA prompt is only one component. As an agent moves from demo to production, the hard problems become specification, state management, access control, and failure design—the familiar territory of software engineering.
ai · agent · software-engineering4 min readEverything worth keeping from a three-hour Lex Fridman podcast: a one-hour prototype that reached 180,000 GitHub stars, a rename war, direct conversations with Zuckerberg and Altman, the philosophy of Soul.md, and why vibe coding is a slur.
AI · open-source · OpenClaw10 min readAn AI assistant ties together my smart home, automation workflows, voice control, and daily information. This is a snapshot of a system that keeps evolving.
ai · homelab · automation6 min readA real attempt to run a multi-agent system on a local 4B model, from compressing the system prompt to giving up—and the capability boundary that attempt exposed.
ai · agent · local-model4 min read
Ariaki