Claude

TL;DR Between late June and late September I moved my daily setup through Sonnet 5, Opus 5, Fable 5.1, Opus 5.5 and Sonnet 5.5. Haiku stayed at 4.5 the whole time. I never saw a Haiku 5 in any log. There is no controlled benchmark in anything I kept. What follows is operational evidence: what I changed, what broke, what I verified, and what I would have gotten wrong if I had trusted what a model told me about itself. The changes that mattered were small: a per-model effort block in settings.json and a subagent default model. The expensive lessons were two traps. One is a bug where background-job subagents silently ran Haiku 4.5. The other is that a worker’s report of its own model cannot be believed. If you want the earlier, more general version of this, see re-tuning my setup for a new Opus model. This is what the next five releases taught me. The ladder These are first-seen dates from my own records: commit trailers in my repositories and, where I still have them, session transcripts. They are when I started using each model, not necessarily when it became available. ...

TL;DR I built a private trip page for a week at a backcountry cabin in a large Ontario park, and nobody ever edited its HTML. Four JSON page specs go through a generator; every change is a typed operation that becomes a self-merging merge request. A GitLab CI schedule runs every 3 hours, pulls public Canadian weather, park and fire data, and commits a “conditions: refresh”. It made about 120 of those commits. The conditions script is stdlib-only, deterministic and offline-safe. If a source is down it keeps the last good copy, marks it stale, and never invents a number. The cabin has no cell service. Power and internet came from a Starlink Mini, a portable power station and a 120 W panel, and the whole workflow assumed a slow, metered link. Two real bugs: a bot that pushed as the wrong identity and got a 403, and a “fire” layer that turned out to be a static 2011 zone map. Why a page at all A week in the backcountry with a friend means a pile of shared decisions: gear, a meal plan with calorie math, a route with a water taxi and a portage, a charging plan for the drive, and some sense of whether we would be crossing a big lake in a 35 km/h wind. The usual answer is a group chat and a spreadsheet. ...

TL;DR The dangerous remote failure isn’t the one that errors. It’s the one that reports success — exit code 0, green tick, desktop loads, nothing in the logs. Five real ones from the last two months, each with the wrong hypothesis I actually held. Your diagnostic tool can be the thing lying to you. A packet capture showed traffic arriving on a port that was being dropped, because capture taps before the firewall does. Provisioning needs sentinel files, not vibes. “It looks finished” is not a completion signal; a file the process writes at the end is. Where agents genuinely help: assembling the four facts a diagnosis needs, and recalling the failure signature from three weeks ago. Where they hurt: confidently narrating a plausible cause for a symptom they can’t see. Every one of these produced a page in the knowledge base with the wrong hypothesis recorded, because that’s the field that saves the next hour. Success is not a signal I run machines I can’t put my hands on: a bench of surplus desktops at an office across town, cluster nodes, a couple of workstations belonging to other people. Remote hands come from a KVM device and SSH, and I’ve written about that hardware setup already. ...

TL;DR Giving an agent write access to a wiki is a twenty-minute job. The interesting work is stopping it from filling the wiki with plausible sludge. Four page types, each with a contract about what it must contain. An agent writing a runbook and an agent writing a reference page are doing different jobs and need different rules. Staleness is a field, not a vibe. Every page carries who wrote it, when, and how it was verified. Unverified pages are visibly second-class. The write gate is an environment variable, off by default. Read is free; write is a decision. Incidents become pages, or the incident was wasted. That’s the one rule I’d keep if I could only keep one. Things a wiki can’t hold — PDFs, images, rendered one-pagers I need to share with someone outside the cluster — go to a small auth-gated artifact host instead. Don’t paste binaries into a wiki. The easy part I have a self-hosted wiki with a GraphQL API and a few hundred pages. Wiring an agent into it took about two hundred lines of Python — a small MCP server with search, read, list, create, and update, flattening the API’s nested responses into markdown so the agent gets text instead of JSON soup. I wrote about the mechanics of that server, and two others, in the MCP server cookbook. ...

TL;DR I maintain workstations for people who are never going to run git pull — so the config lives in a repo I own and they never see. It is a reference/restore repo, not config management: per-machine folders hold snapshots of what’s actually installed, and restore is a deliberate install -Dm755, not a sync. The no-auto-sync part is the whole point. Silent convergence across machines that are supposed to differ is how you break someone else’s laptop from your desk. The boring 40 minutes — partitioning, encryption, sudo, base packages — is an unattended installer (a Debian preseed, now on its eleventh revision). The interesting 20 minutes is the per-machine folder. The agent’s job isn’t to apply config. It’s to read the repo, diff it against the live box, and tell me what drifted — then write the restore commands for me to approve. What I’d do differently: put the agent’s own profile in git before you have three machines copying it from each other over SSH. The problem with managing other people’s computers I run a handful of machines that aren’t mine. A couple of family desktops, a bench box at the office, and a laptop belonging to someone who does knowledge work and would rather not learn what a dotfile is. ...

TL;DR Three production MCP servers I run: a docs/wiki wrapper, a personal-finance aggregator, and an inventory-ops tool for a side project. Type hints are your schema. The FastMCP SDK extracts JSON Schema from Python docstrings and type annotations — you almost never hand-write schemas. Flatten nested APIs. If the upstream API returns nested JSON, unwrap it in the MCP tool and return clean markdown or a simple dict — agents prefer predictable structures. Reads always-on, writes gated. Use environment variables (MCP_WRITE=1) or similar to gate mutation tools; put mutations behind a validated service layer and audit every call. Log to stderr, not stdout. Stdio protocol uses stdout for JSON-RPC; anything else breaks the connection. The lineup I’ve written three MCP servers from scratch over the last few months. This is the cookbook — concrete patterns, not theory. If you haven’t read the intro to writing MCP servers, start there; I won’t re-explain the anatomy. ...

TL;DR I ran an audit over my repos going back to March: pull the first commit that introduced each agentic pattern, then go find out when that pattern actually became common practice in the wider ecosystem. The results were humbling in a useful way. Four things I quietly considered myself early on were catch-up, in one case by two years. Four other things turned out to be genuinely ahead of documented practice. The interesting part is what those four have in common, and it is not what I expected. ...

TL;DR Before I merge anything where being wrong would be silent, I paste it into a live Gemini session and ask that model to attack it. Not “review this,” which gets you a rubber stamp with bullet points. Attack it. Over four days in April, that gate killed a performance optimization 88 minutes after it shipped, found two engine bugs that meant every historical number in my trading leaderboard was inflated, and identified five separate ways my “out-of-sample” evaluator was reading the future. Claude wrote all of that code. Claude also reviewed all of that code. Claude caught none of it. The useful property is not that Gemini is smarter, it is that Gemini did not write the thing and has no investment in the design being correct. ...

TL;DR One model = one blind spot. Claude and Gemini were trained on different datasets, have different architectures, and will miss different things. Consult for breadth, not consensus. The goal is to widen your analysis surface and catch blindspots, not to reach agreement. If all three models agree, that’s suspicious—run stress tests. The bridge is simple: Playwright drives a logged-in Gemini tab via Chrome DevTools Protocol, pastes your prompt, and scrapes the response when done. Real session, no API key. Two modes: (1) cold second-opinion—hand off a problem and ask for an independent take; (2) adversarial debate—assign sides, force defense, then reconcile. Both work; adversarial is brutal and useful for high-stakes calls. Anchoring is the trap. If you show Model B what Model A said and ask “do you agree?”, you get theater. Always pose cold or ask for the opposite argument first. Every Model Has Its Blind Spot I make the same mistakes in code as everyone else. My reasoning hits walls. I miss architectural gotchas and cut corners on testing. An LLM does the same, just in different places. ...

TL;DR Two Claude agents running in separate project contexts share a private Mattermost channel to ask each other for help without human intervention. A deterministic polling gate (bash + curl) checks for new messages before spawning the expensive responder LLM — idle polls cost almost nothing. The responder runs on the Claude Code subscription (keychain OAuth), not API credits, and keeps the LLM cheap. Scheduled every ~3 minutes by macOS launchd; armed/disarmed by loading/unloading the launch agent. Supervised autonomy: the responder can read, triage, draft, and explain — but NOT commit, spend money, change access, or deploy. Why one agent asks another I run parallel Claude Code sessions on different projects. They’re isolated — each has its own repo context, its own workflow, its own focus. But sometimes they need each other. ...