Sora Morimoto

AI Safety. @Cambridge_CL @OCaml_org @TC39. Formerly: @YCombinator (Funded by @JapanGov & @TransposeVC) @Tarides_

[Mistral AI @MistralAI]

Introducing Mistral OCR 4. It creates structure with bounding boxes, block classification, and inline confidence scores in 170 languages. 🧵👇

[Sora Morimoto @smorimoto]

$200のプランにしてても5時間リミットが厳しすぎて大きめのタスク渡すとレートリミットに当たって、ずっとリトライ走ってしまう。
サブスクリプションと別で常に$300くらいクレジット入れておけばある程度は安心という印象…？

[Oikon @oikon48]

Sakana Fugu の $20プランだと、fugu-ultra あんまり動かせなかった...🐡
まともに動かすには、$100プラン以上必要っぽい。

[Sora Morimoto @smorimoto]

As far as I know, the idea of a relocatable OCaml compiler has been around since I was at Tarides, so it's been in the works for years now...
David’s been driving the project from the start, and honestly, he’s an incredibly impressive person.

[Anil Madhavapeddy @avsm]

Also, a multi-year effort to make the OCaml compiler relocatable landed in this tree. Binary caches (uv style) are entirely possible now once a few downstream packages (like ocamlfind) have patches to add support

[Sora Morimoto @smorimoto]

全体的に良いけどData trainingがデフォルトでオプトインになっているので、そこだけ注意かなあ…

[Oikon @oikon48]

Sakana Fuguがリリース👀

Fugu UltraはClaude FableおよびMythosに匹敵するとのこと

https://t.co/rI1Tnf6wIp

[Aiden Bai @aidenybai]

Introducing cnfast

A 3x faster drop-in replacement for `cn`

npx cnfast migrate

[Bull Theory @BullTheoryio]

BREAKING: The NSA's own director says Mythos broke into almost all of its classified systems in hours.

Per The Economist, Senator Mark Warner, vice chair of the Senate Intelligence Committee, said General Joshua Rudd, who runs the NSA and the Pentagon's Cyber Command, told him this directly.

This came out on June 11, the same day Amazon reportedly found a separate jailbreak in Anthropic's models. Within hours, Trump ordered Anthropic to cut off foreign access to Mythos and Fable.

Anthropic shut both down completely instead.

Now there are two competing stories for why this actually happened.

One says the shutdown was a response to the NSA's own classified systems getting breached in hours.

The other says Anthropic is privately pushing back, calling the jailbreak minor and the shutdown an overreaction to something other AI models can already be tricked into doing.

The NSA was already using Mythos for its own cyber operations, with Anthropic engineers embedded inside the agency. The same tool the agency was actively relying on is the one its own director says broke into almost everything it owns.

[今井翔太 / Shota Imai@えるエル @ImAI_Eruel]

「論文はプロンプトになりつつある」というパワーワード。
杉山先生がこれを言うくらいなので、時代を感じる。
https://t.co/Xg9kbj7c6K

[Cloudflare Changelog @CFchangelog]

Wrangler 4.102.0 now supports `--temporary` deployments for AI agents. Skip the manual OAuth flow and let your agent spin up a preview environment directly.

https://t.co/Aopfa32v5f

[R-GRAY @Frozen_Frog_8]

とあるクソガキはこう考える：
俺はMODを作りたい。Javaって何だ？

他のとあるクソガキはこう考える：
レッドストーン回路って凄い面白いな。これってExcelに応用できないか？

また別のクソガキはこう考える：
俺も遊びたい。x86の手軽なデスクトップ環境構築ってどうやるんだ？

更に別のクソガキはこう考える：
俺は友達が24時間遊べるサーバーを提供してやりたい。出来るだけ金はかけたくない。俺に出来ることはなんだ？

15年経ち、彼らは優秀な技術者に育った。WASD以外に押すキーが増えただけの、優秀なクソガキの群れだ。クソガキ共、今日も元気か？俺はそういう「何かやってみたいクソガキ」にPCを売ることでガチャ代を稼いでるクソみたいな大人の一人だよ。

[Jarred Sumner @jarredsumner]

In the next version of Bun

`bun build --react-compiler` runs the React Compiler in Rust

On a large React codebase, it's 19x faster than the Babel plugin

[Tienson Qin @tiensonqin]

Now we can use OCaml to write native iOS/Android/Mac apps with @janestreet's https://t.co/EeeMtKDKkA, can't be more exciting!
@yminsky

[The Wall Street Journal @WSJ]

President Trump said negotiations with Anthropic over restoring access to the company’s latest artificial-intelligence models were “going fine” https://t.co/Xpc9Z08w4p

[SpaceX @SpaceX]

SpaceX has exercised the option to acquire @cursor_ai in an all-stock transaction with the goal of building the world’s most useful AI models.

For the past few months, SpaceXAI has been jointly training a model with Cursor, which will be released in Cursor and Grok Build soon.

We look forward to working closely with the Cursor team to advance our frontier AI capabilities

[Mitchell Hashimoto @mitchellh]

My heuristic is that any diff an agent generates over ~1500 lines is too big and is indicative that the problem needs to be decomposed. This is my general pattern now for feature work:

1. Try to implement the whole feature, loosely guided. I call this the "draw the owl" prompt in reference to the meme. Expect garbage, you're going to get garbage.

2. If the diff is less than 1500 lines, review it and iterate normally. If the diff is more than 1500 lines, prompt the agent to decompose the problem into atomic, incremental, reviewable tasks. Simultaneously, do this yourself.

3. Agents will very often make these tasks way too specific to the shape they solved. You need to massage it into the right general shape. Do that.

4. Kick off new agents to work on those incremental things (as parallelized as possible). Apply the same rules.

5. At a certain, point, repeat the "draw the owl" prompt. At some point, you will get beneath your review-ability threshold.

This has been producing consistently high quality, maintainable, reviewable chunks of code that have a good handoff to either merge as-is or human refinement.

And with the latest frontier models at xhigh thinking, these are all slow enough that you can usually have multiple going concurrently while you are actively reviewing others or working on your own tasks.

HITL (human-in-the-loop) agents are still super important, especially for feature work. Features touch the human boundary in terms of UI, API, etc. And net new stuff can introduce pathologies in the architecture that violate desired invariants (these should be represented in specs or tests but we aren't perfect!).

I know a lot of the leading edge agentic discourse is about "loops" and agents driving agents continuously. I do some of that (will report on that later). But, in terms of raw daily get-shit-done type of work, this is my most rewarding pattern at the moment.

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[Sora Morimoto @smorimoto]

確かにApple Scriptは結構しっかり良くて、macOSの Mailアプリとかのメール本文や、メールの添付ファイルだったりも取れるから活用の幅は広い

[中嶋 謙互 @ringo]

CodexをPMにして、Cursor(Composer 2.5)をコーディング担当にすると、作業がだいぶ速く終わるしトークンもめっちゃ安い。これはすばらしい。。
Codexが Apple Script経由でCursor IDEを叩くのは速いし安定する。 cursor のCLIを使う必要もない。

OSの機能すばらしいなぁ。

[Sergio Paniego @SergioPaniego]

frontier agents are this good partly because the model was trained inside the very harness it ships with

great to see this recipe moving to the open with works like the new "Polar: Agentic RL on Any Harness at Scale" by @NVIDIAAI

it turns harnesses (codex, claude code, qwen code or pi) into RL training environments without touching their internals

[Brayden @BraydenWilmoth]

Almost felt like a launch week at Cloudflare...

- Partnered with @xai to bring Grok to Cloudflare
- OpenAI announces Sites built with Cloudflare
- Acquisition of @voidzerodev & Vite team
- AI Gateway spend limits released
- Now support self-managed OAuth clients

🥵

[RaceJohnson @RaceJohnson]

discomorphism

[KC Sivaramakrishnan @kc_srk]

I'm offering "Functional Programming with OCaml" on the NPTEL platform in July 2026 sem. Enrollment is open now.

The first 8 modules of the interactive book should be fairly stable. The rest is still in development. Sharing early in the spirit of building in the open.