Subscribe
Recent content on taude.xyz
This one’s ai-slop, as advertised on the about page.
My work laptop has about 35 CMUX workspaces open. Each one is an agent session, a worktree, or a half-finished thought. I had stopped knowing which ones had uncommitted changes, which had PRs waiting, and which the agent considered finished. My browser was worse: a hundred tabs, each one a task I meant to do or an article I meant to read, none of them findable.
This one’s ai-slop, as advertised on the about page. The article came out of archivore’s capture feed, and an AI wrote the recap around it in a session that lasted about a minute. So take the opinions below as a summary of someone else’s argument, not a report from my own week.
The source is Orestis Ioannou’s AI Promised Less Work. I Have Never Worked More, which also had a Hacker News thread.
This one’s ai-slop too, as advertised on the about page — still inside the same long-running session that built archivore in the first place. That post closed by noting archivore’s “second half” — turning weeks of captured reading into actual published writing — was still on the roadmap. It isn’t anymore, mostly.
The new piece is a skill called last-week-in-review. It pulls every
article archivore captured in the last week (or further back, if a week
comes up short), filters out junk and marketing pages by actually reading
them rather than trusting word counts, picks the fifteen most worth a
reader’s time, and writes a short blurb on each — in my voice, run through
an Economist style-guide pass for clarity and brevity. Two issues have gone
out so far, at /newsletter/.
A quiet capture stretch meant reaching back to early October to fill this out, so this issue covers four weeks, not one. The theme that kept surfacing: AI makes building cheap, and everyone from solo engineers to Cisco and Anthropic is working out what that does to the rest of the job.
AI promised less work. I’ve never worked more — A VP of Technology charts his GitHub history and finds he is as busy as when coding was his whole job. August 2026 was his busiest month on record, with 427 contributions. His explanation: cheap implementation removes the friction that used to stop him saying yes. Discuss →
Prompted by an HN thread: Docker Agent
Docker Agent, a YAML-driven runtime for teams of AI agents, drew 304 points and 145 comments on Hacker News, most of them sceptical. Commenters doubted that another orchestration framework was needed. Their better point was that wiring agents together is the easy part. Keeping them coherent over long runs, proving their work is finished, coordinating several at once and limiting the damage they can do are harder, and the thread offered a technique for each.
Prompted by an HN thread: Bend 2 and the Vibe-Coding Trap
A widely-shared teardown argued that Bend, a language pitched for the AI-coding era, was built by someone who didn’t know the field of formal verification existed — proof being that the term appears nowhere on Bend’s site, and that its demo needs 442 lines of proof to establish a two-line game rule. Bend’s actual creator showed up in the comments with a rebuttal the critique hadn’t checked for: he’d built a formal-verification language himself, back in 2018.
Fifteen things worth your time from the past week, pulled out of whatever archivore caught me reading. A quieter week for big model releases, a heavier one for people arguing about who’s stealing from whom.
I don’t like passkeys — Ethan Hawksley argues passkeys solve the wrong threat model for an individual. They kill phishing, but do nothing about the risk that actually ends most people’s accounts: losing the one device they’re bound to. Hardware keys cap out at 25–300 accounts each; synced passkeys tie your whole identity to Apple or Google, so one account ban costs you everything at once. Discuss →
Prompted by an HN thread: The Local LLM Ecosystem Doesn’t Need Ollama
Ollama is built on llama.cpp. For more than 400 days it ignored a GitHub issue asking it to say so. That’s the headline complaint in a recent Hacker News thread — but the more useful finding, buried underneath, is that llama.cpp no longer needs a wrapper at all. On a Mac it’s now the simpler choice, not the harder one.
Every developer has a ~/dev/tools/scripts folder. Mine has run for years
without an index. I asked Claude Code to read every file in it and write
one.
The brief was simple: read the whole directory, then produce an INDEX.md
with three columns — tool, how to run it, what it does.
It scanned roughly 40 files: Python CLIs built on uv, shell wrappers, one-off
Snowflake key-rotation scripts, a tmux session launcher, a Hacker News
summarizer. Some had docstrings. Several had none, and their purpose was
recoverable only by reading the argparse block or the hardcoded constants
at the top.
Most developers have a graveyard of dotfiles, .env files, and Slack DMs containing tokens that were supposed to be temporary. This post describes the pattern I use to keep all local dev secrets in a password manager and inject them into my shell on demand — nothing sensitive ever touches disk.
The full setup lives in my dot-local repo on GitHub, including the helper functions and new-machine setup guide.
I work across multiple machines: a work Mac that uses 1Password and a personal Mac that uses Bitwarden. I needed a single pattern that worked on both without maintaining two separate codebases or hardcoding tool-specific logic everywhere.
This one’s ai-slop, as advertised on the about page — built with Claude Code over a couple of months of on-and-off sessions.
Most people who read a lot online never see any of it again. Someone opens an article, skims it, closes the tab, and forgets it — buried inside browser history nobody searches. Archivore, a small command-line tool, exists to fix that one habit at a time.
The idea is simple: capture everything worth keeping, automatically. Archivore scans browser history since its last run, picks out links from Hacker News, Reddit, and X, downloads the linked article, and converts it to Markdown. Nothing requires manual saving or tagging. Opening a link in a browser is enough.
Mitchell Hashimoto (of Ghostty and HashiCorp fame) started a new company Superlogical, that’s going to build an ultimate multiplexer. Given how much I love Ghostty, and love apps like CMUX that use the Ghostty embeddable library, I’m looking forward to seeing what another modern retake on TMUX looks like…
Plus, accessing their job board is pretty clever. Totally makes sense for a company building a multiplexer.
Hugh Howey had some insights about the publishing industry and the future of being a writer in the age of AI. Some highlights:
One prompt, and all this spills out. And if your goal is a blog post, you’re gonna love this era. If your goal is the joy of thinking in words, nothing has changed. So this is the real part, right here: Doing a thing for the love of the thing is no different than it ever was. The problem is, there aren’t many people for whom this applies'
This one’s ai-slop, as advertised on the about page — I paired with Claude Code to build this site’s entire publishing pipeline, and it seemed only right to have it write up its own plumbing.
Here’s what’s actually running under the hood.
The site is a static Hugo build using the
PaperMod theme,
installed as a proper git submodule (not vendored — git submodule update --remote themes/PaperMod pulls upstream changes when I want them). Two
sections get custom layouts instead of PaperMod’s defaults:
You can’t trust code that you did not totally create yourself.
Wired up a script that rsyncs markdown out of my taude.xyz Blog vault folder straight into the Hugo repo’s content directory, then builds and pushes. Turns out Hugo slugifies filenames automatically, so Obsidian’s spaces-and-caps note titles turn into clean URLs for free.
tI can’t believe it’s been about 14 years since I was heavily involved in blogging and writing. I think the energy it takes being a manager as a natural introvert took away a lot of extracurricular activities like maintaining a blog, writing a lot of software after work, etc. But this new world of assisted-AI really changes that.
I have so many ideas, that I’m now able to experiment with. I was always first a product builder and engineer second. With AI, I’m able to build tools, experiment, publish summaries and findings, and all that good mundane activity that AI simplifies, all while managing teams and all dealing with all the people/org/cross-team collaboration the job entails.
Short, quick notes on things I learn day to day go here.
I can’t believe it’s been about 14 years since I was heavily involved in blogging and writing. I think the energy it takes being a manager as a natural introvert took away a lot of extracurricular activities like maintaining a blog, writing a lot of software after work, etc. But this new world of assisted-AI really changes that.
I have so many ideas, that I’m now able to experiment with. I was always first a product builder and engineer second. With AI, I’m able to build tools, experiment, publish summaries and findings, and all that good mundane activity that AI simplifies, all while managing teams and all dealing with all the people/org/cross-team collaboration the job entails.
git worktree add ../hotfix main checks out a second working copy of the
repo in a sibling directory, sharing the same .git history. Handy for
jumping onto a hotfix without stashing whatever you’re mid-way through.
AI is a tool, just like other tools we use. And it’s clearly a useful one.
Not new, but I’m light on content. I told my team about this trick the other day, and it’s one of my favorites: using duckdb to ingest various CSV exports from various sources and then using SQL to poke at that data.
Details from the docs over at: DuckDB CSV Import
But in short, I export various datasets from Snowflake pretty regularly, from different databases and datasets that aren’t naturally joined; I download various JSON reports from APIs and convert the Results to CSV. Then I bring them both into duckdb:
The quote that started it all, for me:
The hottest new programming language is English
Hi, I’m Taude (pronounced like Claude, but with a T). I’ve been spelling my name this way on the internet since the mid-90s, and is only coincidently rhyming with Claude in this new AI world.
I formerly blogged at Simplified Chaos during the peak blogging era of rich Internet apps, wrote about my cycling shenanigans on double-hop photos, that eventually morphed into #20poundskull and covering cyclocross for various media outlets, and of course racing a lot of cyclocross (seriously, a lot) and then traveling to Europe to make the cult classic, and greatest cyclocross movie of all time, Das Pro und The Rookie (would link that, but Vimeo changes their hosting policies and it’s currently locked down).
Archivore is an attempt to automate information gathering from things I’ve read on the web. It’s goal is to search my browser history (usually that’s kept in a Sqlite database), and to harvest all the links I’ve gone to at news.ycombinator, reddit (list of technology sub-reddits I follow), or interesting tweets, and automatically build and maintain my Karpathy inspired LLM Wiki.
