Developer News

視覚皮質を再学習させるVRアーキテクチャ：AmblyoPunchの設計思想 最近、SanDiskがMetaの次世代VRグラス向けに公式の拡張ストレージを提供することが発表されました。スタンドアロン型ヘッドセットにおけるI/Oおよびストレージ容量の制限は、これまで高負荷なシミュレーションや詳細なユーザーログを伴う専門的アプリの展開において大きな壁となっていました。このハードウェア的な制約の解消は、リハビリテーションや教育といった「専門的トレーニング」のあり方を根本から変える可能性があります。 1.

Less than two weeks after its $11,000 headline, the configurable deep learning workstation has dropped every Nvidia option and the tinybox pro, added two Radeon AI Pro R9700 builds, and raised their price.

I've been building a predictive coding (PC) engine from scratch in C++20, with a thin Python layer on top, and just cut the first public release. Posting here for feedback and because I haven't seen many from-scratch PC implementations with real kern

AI is scaling faster than almost any technology before it. But behind every intelligent response lies a costly infrastructure, an uncertain business model, and a question every developer should be asking: Who can afford to build the future? Artificia

If you have written concurrent code with threads, you know the routine. Several threads touch the same data, so you put a lock around it. Then you add a second lock, take the two in the wrong order on a busy day, and two threads wait for each other f

Comments

TL;DR - Teams treat SFT, preference optimization, and distillation as swappable, independently-testable modules in a post-training pipeline. They aren't. The order you apply them changes the final model in ways that per-step evaluation never catches,

A web page can save an AGENTS.md into your Downloads folder without a single click. Your coding agent reads it as project rules. Here's what happened, what we saw in our own test, and the 2-minute check. Your Downloads folder is the only folder on yo

While everyone watched LiteLLM, a quieter campaign spent five weeks doing something worse: taking stolen GitHub tokens and committing a secret-dumping workflow into every repository they could reach - 13,500 planted branches across 850+ compromised a

This is a submission for the Hacktoberfest Open-Source AI Challenge: Slow and Steady . What I built If you develop with Flutter, you may know the frustration of spending half an hour debugging an error, fixing it, and then meeting the same problem ag

Thirty years of confident wrong answers I have spent thirty years walking into rooms where a distributed system was misbehaving in production and somebody was certain they knew why. Often they were wrong. Not stupid. Wrong, and confident. I went to s

This is a submission for the Hacktoberfest Open-Source AI Challenge: Week 4 Last week I introduced SkillSentry - a local-first security scanner for AI agent skills ( SKILL.md files) and MCP server configs. Its whole philosophy is open by design: stat

A configuration mismatch in GLM 5.3 on Workers AI triggered an SGLang bug that exposed KV cache pages across requests. Here is how we addressed the issue.

テクノロジーの安全性をどう担保するか──OpenAI提訴と、弱視VR『Amblyotube』の二眼独立レンダリングが示す「誠実な特化」の設計思想 OpenAI が提訴されたニュースを目にしたとき、私は自分のプロジェクトでも「安全性」をどう定義し、どう実装するかを改めて問い直した。 当時、娘が弱視（amblyopia）と診断され、眼科医から「両眼視機能を鍛えるには、左右の目に別々の映像を提示し、脳が融合させる練習が必要だ」と言われた。市販の VR ヘッドセットは便利だが、一般的なレンダリングパイプ

Anthropicの黒字化が示す「実用AI」の波、VR弱視トレーニングにも届いていた話。 AnthropicがQ3で売上168億ドル、調整後ベースで黒字化したニュースは、AIスタートアップが「実用段階」へ一気にシフトした証左だ。同じ波がヘルステックにも押し寄せ、私はVRエンジニアとして、不同視（弱視）を持つ娘のために何かできないかと模索していた。そこで出会ったのがMeta Quest向け『Amblyotube』だ。 アプリの核心は、YouTube動画を再生しながら左右の目に別々の映像を提示するd

Disclosure: I built CloudHike , so read this with that in mind. Supabase details come from Supabase's own docs and pricing pages and were checked on October 26, 2026. Your coding agent can write a whole app in an afternoon. Then it hits the backend a

An agent turn is not a message. It is the whole conversation so far, re-sent. I spent a week logging per-turn usage objects off the OpenAI-compatible endpoint my agent talks to, and the thing that surprised me was not the total - it was the ratio. Pr

If you use coding agents across several projects, you probably maintain the same instructions in several places. A rule gets copied into CLAUDE.md in one repo and AGENTS.md in another. The copies drift . One gets fixed, the others don't. Important ar

A build container shouldn't need a pretend monitor to compile a stylesheet. Until now, parts of our CSS build path pulled in desktop machinery that made that a practical concern. What is Codename One? Codename One is an open-source framework for buil

My AI agent can write a whole Next.js app in an afternoon. Then it stops and hands the hard part back to me: make a Postgres database, set up a storage bucket, find a Redis instance, add auth, pick an email provider, and copy a dozen connection strin

In 2024 the skill everyone talked about was prompting: find the right words and the model does the right thing. In 2025 it was context: hand the model the right files and docs, and the exact wording matters less. In 2026, when an agent fails, it is r

The photo at the top is night one: a TV on the living-room floor showing the Ubuntu installer, and the little PC with the blue glow that now renders every film on kitcut.ai. TL;DR: our AI video studio went from an overheating laptop, to a slow $250/m

I run language models on my own hardware, and the hard part was never running them. It was living with them. A terminal for quick questions, a separate web UI per server, history scattered everywhere. So I built Klepai: a free, open-source desktop ap

Introduction Kubernetes supports more than continuously running application workloads. It also provides controllers for node-level agents, one-off batch processing, and scheduled tasks. In this mission, I implemented a DaemonSet to run a lightweight

Salesforce is preparing an important change in how artificial intelligence agents interact with its platform. And I believe everyone working in the Salesforce ecosystem should be paying attention. Not just those using Agentforce. I'm also talking abo