To main content To menu

Twelve months, unabridged The year 2026, start to finish

65 posts, newest first. A few have aged in public.
A narrow Sài Gòn residential alley at night, seen down its length, with a lone figure walking away between pools of orange lamplight, tangled power cables overhead and roller gates closed on both sides

Six Years Of Alleys Somebody always opened the gate

Safety in Sài Gòn is not ambient - it is made of people. Six years of hẻm life across four districts, and what Phú Mỹ Hưng costs.

Spent a few days with an abliterated Qwen3.8-27B on the M4 Pro. It answers everything. Bombs, malware, whatever you feed it, no hesitation. It also hallucinates like it gets paid per assertion, which makes sense once you notice what abliteration actually removes: the model’s capacity to say no.

There is research on this. Strip the refusal direction and MMLU, HellaSwag and IFEval all stay within a point. The only real regression is TruthfulQA, down 7.1.

Now read the model card. MMLU, MMLU-Pro, GSM8K, CMMLU, perplexity. No truthfulness benchmark anywhere.

The one thing that breaks is the one thing nobody measures. Released strictly for legitimate research, says the disclaimer sitting above the buy button.

A wooden serving board held out toward the viewer with a round golden carrot cake on it, one wedge cut out showing the orange-flecked crumb, an open-plan office with concrete ceiling and ducts behind

Public Alpha, Edible The carrot cake I shipped to production untested

Carrot cake with no butter and no oil, baked at 11pm from memory alone and served to an entire office before anyone had tasted it.
A small tethered hot air balloon drifting in layers of mist above a hilly jungle canopy, with the outlines of an overgrown stepped pyramid and broken ruins visible through the gaps below

On Repeat, Regrettably A stranger in Hamburg rebuilt the Tomb Raider theme, twice

Tomb Raider theme, covered twice by German producer Ambyion - a 2021 remix, a 2025 rework, and one loop I cannot get out of.

Four language models, four Age of Empires 2 strategy scripts, from scratch, no peeking at each other. They failed the same way: hoard villagers, send hunting parties halfway across the map, never bank enough to advance an age. Gemini never left the Dark Age. Four vendors, one set of bad instincts.

The real find is what Emergent Garden built after: an overnight loop where a model mutates the best script, runs a tournament against the previous winner and the built-in extreme AI, keeps what survives. Genetic programming with an LLM as the mutation operator instead of a coin flip. ≈15 USD of tokens, and it beat extreme – by a hair, with a bot he then stomps by walking cavalry around the back.

Commenters point at the ceiling: the game ships Direct Unit Control – loops, pointers, near-human micro – and the models barely used it.

Illustration of a teenage protest march passing along a city boulevard in cold spring light, one marcher holding up a piece of brown cardboard painted with three eggs

Eggs, Beats And Paragraph 188 Merz leck Eier, or how a confiscated sign became a banger

Merz leck Eier went from a confiscated demo sign to a viral rap single, a Google Maps prank and an open prosecutor file - here is the story.
A long candlelit banquet table in a high hall, seventeenth-century figures leaning over an open ledger, a potted palm behind them, and one lone cloaked figure waiting outside in the doorway

Words Nobody Asked For A kennel for maidens

German was argued into existence by people who mostly lost - four centuries of poets, pedants and purists building a language out loud.
Low-poly illustration of impossible glowing green rock shards floating in a black void, with an ochre island carrying a red pagoda temple, a stone lantern and a blossom tree, and a small figure with a braid leaping between them under a pale moon

Thirteen Years Later The better tomb, and the man who wouldn’t build it

Tomb Raider 2 is the better, rounder Tomb Raider - and the game its own creator refused to make. On levels, gunplay and the Lara hype machine.
Ligne-claire illustration of a bearded island hermit in a rainbow robe with a red headband, standing on a grassy headland above a turquoise lagoon among tall seven-leafed plants, a wrecked rowboat on the beach behind him

Twenty-Five Megabytes Of Island Stranded 2 still eats weekends

Stranded 2 runs on Linux in 2026 - a 2007 freeware survival island, a lost weekend, and one German guy still building the sequel.

For a couple of months now every terminal I own wears the same bar: coralline, a Powerlevel10k-inspired statusline for Claude Code. All segments on, tokyo-night, two lines. Branch and dirty state, the active model plus its effort level, the context window filling up, the 5h and 7d rate-limit gauges with their reset countdowns, session cost, session length. I have stopped guessing how close I am to a wall.

Run the installer yourself, or hand the playbook to Claude Code and let it interview you. The readme’s own trust section says a Claude that stops to inspect that playbook first is behaving correctly. It is. Read the script, then run it – trust is good, verification is better.

Needs jq and a Nerd Font, or VL_ASCII=1 without the font. Renders locally: no network calls, no tokens.