Solving autonomous Harness building ex-@ZyphraAI
NL/NO 🇳🇱🇳🇴
Joined March 2011
- Tweets5.6K
- Following535
- Followers2.3K
- Likes17.1K
> secure
> RCS
By popular request, you can now text your dot.
Give it a task to take off your plate or something to keep an eye on, using text, photos, or even voice notes. It will text you proactively with relevant updates.
Safe, secure texting via iMessage or RCS. You’re the only one who can text your dot, and your dot can only text you. Rolling out now as a preview for Pro users in the US.
Here's the viral video from the Dutch House of Representatives with English subtitles
#vragenuur
Ja, waarom nodigen we Anthropic niet uit om hier naar de beurs te gaan, zodat wij belasting kunnen heffen en vertellen wat ze wel en niet mogen.
Ik ken haar niet, maar dat moet #D66 zijn.
Anthropic ziet ze al aankomen 😂😂😂😂😂😂😂😂😂🤣
Set the Bash tool's timeout to 290 seconds for every Claude Code config directory on this machine that does not have it yet, so that a command (a test run, a gate, a push) runs in the foreground for at most 290 s and is then moved to the background automatically, where it keeps running and reports back when it ends. Today the maximum is 10 minutes.
Do it in one pass, without asking:
1. Find the config directories: `~/.claude` and every `~/.claude-account*` (and any other directory a `CLAUDE_CONFIG_DIR` on this machine points at). For each, the user settings file is `<dir>/
settings.json`.
2. Skip a file whose `env` already holds `BASH_DEFAULT_TIMEOUT_MS` and `BASH_MAX_TIMEOUT_MS` at `"290000"`.
3. For every other file, merge exactly these two keys into its `env` object, creating `env` if missing, and change nothing else (hooks, permissions, model and the rest stay byte-for-byte the sa
me in meaning):
```json
"env": { "BASH_DEFAULT_TIMEOUT_MS": "290000", "BASH_MAX_TIMEOUT_MS": "290000" }
```
Use `jq` into a temporary file, check it with `jq -e .`, then move it over the original. Where no `settings.json` exists, create one holding only that `env` block. Do not add any other variable: these two are Claude Code's documented settings, and nothing else is needed.
4. Verify each file: `jq -c .env <file>` shows both keys, and `diff <(jq -S 'del(.env.BASH_DEFAULT_TIMEOUT_MS, .env.BASH_MAX_TIMEOUT_MS)' <backup>) <(jq -S 'del(.env.BASH_DEFAULT_TIMEOUT_MS, .e
nv.BASH_MAX_TIMEOUT_MS)' <file>)` is empty. Keep a backup beside each file while you work, and move the backups to the Trash (`/usr/bin/trash` on macOS) once every file checks out; never delete
permanently.
5. Prove the behaviour in your own session: run `echo "$BASH_DEFAULT_TIMEOUT_MS $BASH_MAX_TIMEOUT_MS"; sleep 8; echo done` with the Bash tool's `timeout` set to 5000. It should print `290000 290000` (if it prints nothing, the session predates the change: say so, as a restart picks it up) and Claude Code should report that the command was moved to the background, then notify you that
it completed with `done`.
Report one line per config directory (changed, already set, or created), and the result of step 5. Never open, print or copy any `.env` file or credential while doing this.
Screenless M1 Airs are most cost-efficient to run things on, $30 per 1k multi-core points 8k total GeekBench 6
Servers are like $100-200 i'd guess
Stijn retweeted
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
no autoresearch 🤯
The bottlenecks in AI are moving beyond the models.
We’re excited about startups building the hardware, infra, financial systems, and specialized software to bring AI into the real economy, from factories and data centers to government and space.
ycombinator.com/rfs
Stijn retweeted
Ghostty 1.4 is now out! Full Kitty graphics with animation (the first terminal besides Kitty to do it), scrollback compression for much lower memory usage, macOS 27 support, a new OpenGL renderer on Linux, ghostty +ssh, big performance gains, and hundreds more changes.
The highlights:
- Scrollback compression: scrollback compresses ~90+% on average, meaning your terminals use that much less live memory
- Huge memory savings across the board: no single large thing, just lots of good improvements lowers memory usage everywhere, netting anywhere from 50 to 90% less memory usage depending on workload
- Kitty graphics is now 100% complete, including animation and relative placements, after a full conformance pass against the spec and Kitty itself
- The Kitty clipboard protocol, with an image preview when a program asks to read your clipboard
- Split drag and drop on Linux, including between windows
- ghostty +ssh installs terminfo on remote hosts automatically, so things just work over SSH
- Lots of terminal emulation accuracy fixes
This release is 7 months of work from 172 contributors over 2,889 commits. It also fixes 10 security advisories, so please update.
Full release notes: ghostty.org/docs/install/rel…
woah wtf, TIL, that's why the locals call it a Sherpa stair!
youtube.com/watch?v=vrEZa-HM…
