AIToday

Today's AI, for Today's work.

Start free60,000+ monthly readers

Latest AI News

Earlier

LeapAI automates Trustford intake with Claude Code

LeapAI released the first 'AI Consultant on Tour' episode on September 30, 2026, showing CEO Yuki Iwata automa…

Top Companies AIOct 5, 2026

Reflection AI releases Beam, its first open-source model

Reflection AI released Beam, its first open-source model

Semafor TechOct 5, 2026Also reported by Fortune AI and 3 others

llm-anthropic 0.30 adds refresh and token count commands

llm-anthropic 0.30 ships alongside Claude Sonnet 5.5 and adds llm anthropic refresh, which pulls Anthropic's m…

Simon Willison's WeblogOct 5, 2026

SageMaker AI adds aws-ai-ml skill for coding agents

Amazon SageMaker AI introduced the aws-ai-ml skill via the Agent Toolkit for AWS, plugging into Kiro, Claude C…

Amazon AI BlogOct 5, 2026

Claude Opus 5.5, Sonnet 5.5 land on Amazon Bedrock in AWS GovCloud (US)

Anthropic's Claude Opus 5.5 and Claude Sonnet 5.5 are now available on Amazon Bedrock in AWS GovCloud (US), wi…

Amazon AI BlogOct 5, 2026

GitHub opens ReviewBench, a 219-PR code review benchmark

GitHub released ReviewBench, a benchmark of 219 public pull requests across 19 languages, built by analyzing 1…

GitHub Copilot BlogOct 5, 2026Also reported by GitHub Blog (AI)

Vercel's Python SDK adds evaluate() for Jev

Vercel shipped the latest AI SDK for Python, installable with uv add ai, adding an experimental evaluate() API…

Vercel AI BlogOct 5, 2026

Amazon Bedrock Managed Knowledge Bases: agentic hits 6 of 6

AWS built a LangChain RAG app on Amazon Bedrock Managed Knowledge Bases and ran one six-intent question throug…

Amazon AI BlogOct 5, 2026

aiUsage CLI turns Claude Code and Codex logs into CSV

The Rust-built aiUsage CLI aggregates Claude Code and Codex local logs into usage.csv, one row per model, cove…

Zenn AI/MLOct 5, 2026

JEV sorts n8n inquiries by meaning, not keywords

With JEV plugged into a self-hosted n8n workflow via the TypeSafe AI community node, "I want to move my day to…

Zenn AI/MLOct 5, 2026

AI Orchestra: Claude Code writes SEO column, 3 layers guard quality

AI Orchestra says its column is written by Claude Code, not people, and since starting SEO work in July 2026 i…

Zenn AI/MLOct 5, 2026

16-year hardware engineer: Claude Code takes the paperwork, humans keep the hands

A 16-year hardware engineer says that after months using Claude Code, research, spec drafts, test sequencing a…

Zenn AI/MLOct 5, 2026

DSL wins CAD bracket test 10/15 where code generation hits 4/15

In a FreeCAD 1.0.2 bracket task run 15 times per method with gemini-3.5-flash, the DSL method passed final ver…

Zenn AI/MLOct 5, 2026

TypeScript becomes the pick for AI apps

The article argues TypeScript is the leading choice for building AI-integrated apps — chat UIs, AI agents and…

Zenn AI/MLOct 5, 2026

Thariq Shihipar: CLAUDE.md likely to disappear

On the Latent Space podcast, Thariq Shihipar of Anthropic's Claude Code team said CLAUDE.md will likely disapp…

Zenn AI/MLOct 5, 2026

ML.GENERATE_TEXT vs AI.GENERATE: model objects no longer needed

A Zenn series explains that ML.GENERATE_TEXT needs a model object created via CREATE MODEL and returns a row s…

Zenn AI/MLOct 5, 2026

AgentCore Browser fails on CodeZip, fixed via /tmp copy

Testing Amazon Bedrock AgentCore Browser under a CodeZip runtime returned "PermissionError: [Errno 13] Permiss…

Zenn AI/MLOct 5, 2026

Qwen3.6-35B-A3B picked for a local Claude Code replacement

A developer whose Claude Code subscription ran out of tokens set up a local LLM on his RTX 5070 Ti (VRAM 16GB)…

Zenn AI/MLOct 5, 2026

DeepSeek Harness 0.2 ships with Windows, macOS desktop apps

DeepSeek released DeepSeek Harness version 0.2 on 2026年10月1日, adding desktop apps for Windows and macOS to its…

GIGAZINE AIOct 5, 2026

Solo dev makes /code-review an unskippable delivery gate

A one-person contractor wired Claude Code's /code-review into a pre-delivery script, and review execution went…

Zenn AI/MLOct 5, 2026

Token Dashboard puts Cursor and Claude Code usage on a 9.2-inch side screen

The author built Token Dashboard on a 9.2-inch TURZX USB smart display with a 1920×462 px bar screen, showing…

Zenn AI/MLOct 5, 2026

Claude Code sums up daily logs for a 13-agent operator

A developer running 13 AI agents built a Python script that reads yesterday's freeform daily report file, has…

Zenn AI/MLOct 5, 2026

Claude Code's 25.5k token fixed cost — kou trims Memory to 4.2k

Gakken LEAP engineer kou ran /context in a new Claude Code session and found 25.5k tokens already loaded befor…

Zenn AI/MLOct 5, 2026

Not-yet-AI engineers hold the advantage, writes 20-year SE

A systems engineer with over 20 years of experience, who has run more than a dozen projects in the past half y…

Zenn AI/MLOct 5, 2026

Cognizant's Devin use cuts release time, 37% saving

Cognizant reported that using Cognition's Devin, an AI software engineer, on Odyssey Logistics' systems redeve…

Yahoo Finance AIOct 4, 2026

State of Devs 2026: half of devs let AI write 75%+ of code

Devographics surveyed 5,463 developers from July 5 to September 5, 2026, and found 49% say AI writes 75% or mo…

PublickeyOct 4, 2026

Google's RRSI lifts unseen-task scores by up to 4.7 points

Google researchers' RRSI caps and shrinks how many edits a self-improving agent can make, and a critic rejects…

THE DECODEROct 4, 2026

Remote Control blocks Claude Code auto-continue after limit reset

The author traced Claude Code's auto-continue after a usage-limit reset to Remote Control sessions, reporting…

Zenn AI/MLOct 4, 2026

Python and Claude Code, zero CDN: one brake on AI slop

The author builds a small-business site with Python and Claude Code only — no CDN, no npm, no JS framework — a…

Zenn AI/MLOct 4, 2026

BigQuery's AI.FORECAST adds TimesFM zero-shot forecasting

Google Research's TimesFM foundation model now powers BigQuery's AI.FORECAST, which returns forecasts from one…

Zenn AI/MLOct 4, 2026

Ariadne repair theorem: fixes cost 2 observations, not 0

Kenya Kondo's Ariadne Boundary Repair Theorem, illustrated on a three-value model, shows repair is possible wi…

Zenn AI/MLOct 4, 2026

persona-feedback hits v0.2.0 with per-persona browsers

The persona-feedback Claude Code plugin reached v0.2.0

Zenn AI/MLOct 4, 2026

lossless-compaction cuts Claude Code compaction from 51.8秒 to 0.26秒

A developer released lossless-compaction, a Claude Code plugin that writes no model summary and instead moves…

Zenn AI/MLOct 4, 2026

GitHub, VS Code set rules for AI coding tasks

GitHub's material for its Copilot cloud agent urges developers to state the problem, completion conditions and…

Zenn AI/MLOct 4, 2026

Minecraft AI bot plan hits 月数円 with Jev, Gemini, Mineflayer

The developer published fk2000/minecraft-ai-bot, which uses Jev to assemble instructions, Gemini to pick from…

Zenn AI/MLOct 4, 2026

AI Orchestra's Miyatchi: hand AI agents 3 design decisions, not code

Writing in a Zenn article, Miyatchi of AI Orchestra says the spec for an automated task is just the current hu…

Zenn AI/MLOct 4, 2026

Turn today's AI news into today's work.

The day's essentials from 100+ sources, delivered to Email, LINE, or Slack. Free to start.

60,000+ monthly readers
Get Started Free

Anthropic's Thariq Shihipar: CLAUDE.md will fade

In a Latent Space episode, Anthropic Claude Code team engineer Thariq Shihipar answered a host's question by c…

Zenn AI/MLOct 4, 2026

ELN workflow turns discipline into 165 Claude Code gates

EarthLink Network's ELN workflow packages its development rules as 165 Claude Code skills (AI procedure sheets…

Zenn AI/MLOct 4, 2026

chaffjs + LLM: 5 周 to humanize AI-written Japanese

The author ran an AI-generated technical post through chaffjs and an LLM for 5 周 of edits, cutting 98 lines an…

Zenn AI/MLOct 4, 2026

AI駆動開発 spans 設計 to テスト, not just code

A Zenn article defines AI-driven development as embedding generative AI into 要件整理, 設計, 実装, デバッグ, テスト, レビュー and…

Zenn AI/MLOct 4, 2026

Anthropic adds Wrap-Up Allowance to Claude Code

Anthropic's "Wrap-Up Allowance" lets Claude Code keep going briefly after the 5-hour limit, showing "Usage lim…

Zenn AI/MLOct 4, 2026

Claude Code auto mode alone won't tame loops, CoWorker lead says

CoWorker security tech lead Ito described running CI-watch, parallel-review and overnight loops in Claude Code…

Zenn AI/MLOct 4, 2026

Cameo proxy puts cheap models inside Claude Code

A developer released cameo, a roughly 400-line Go program that reads a TOML file, starts a local HTTP proxy, a…

Zenn AI/MLOct 4, 2026

PyTorch and Arduino UNO power an end-to-end self-driving toy car

The author built a palm-sized car where a PC runs a PyTorch CNN (NVIDIA PilotNet, shrunk) that maps camera ima…

Qiita 機械学習Oct 3, 2026

CS146S notes: three MCP mistakes fixed, Server never calls APIs

Writing up Stanford CS146S course material, the author corrected three of his own MCP mistakes: the Server nev…

Zenn AI/MLOct 3, 2026

Claude Code hooks turn prompt requests into hard rules

A hands-on guide shows three Claude Code hooks — PostToolUse auto-formats edited files with prettier, ruff, or…

Zenn AI/MLOct 3, 2026

ttokunaga: $17,239.93 in AI use cost 36,300円

Developer ttokunaga reported September 2026 usage of about 41.574 billion Codex tokens and roughly 2.098 billi…

Zenn AI/MLOct 3, 2026

AI Security Lab opens Ollama-based LLM test setup on GitHub

The AI Security Lab outlined its standard experiment setup: Python programs call Ollama's API directly to run…

Zenn AI/MLOct 3, 2026

MCP drops sessions in 2026-07-28 spec, David Soria Parra leads

David Soria Parra led the 2026-07-28 Model Context Protocol spec, which deletes the initialize/initialized han…

Zenn AI/MLOct 3, 2026

AIエージェント pipeline: the real bottleneck is queue time

Operating an AIエージェント pipeline from GitHub Issue to pull request, the author found waiting time splits into fo…

Zenn AI/MLOct 3, 2026

Anthropic ships Mods to let devs rewrite Claude Code

Anthropic released "Mods" for Claude Code, JavaScript or TypeScript functions that hook into events like tool…

THE DECODEROct 3, 2026

LEGO-Anything's LEGO-Bench: GPT-6 Astra hits 53.4%

LEGO-Anything has a coding agent write executable Blender code from a single photo, and in the accompanying LE…

THE DECODEROct 3, 2026

Ops Console gives ALPHA FORGE a 3D cockpit for its four AI models

A solo developer of ALPHA FORGE built an "Ops Console" on fully separate Next.js 3003 / FastAPI 8003 processes…

Qiita 機械学習Oct 3, 2026

Synopsys ties EDA, IP revenue to customer output at 2026 Investor Day

Synopsys said at its 2026 Investor Day in New York that it is reshaping how it monetizes electronic design aut…

DIGITIMES AsiaOct 3, 2026

Python mean-reversion tutorial uses yfinance, SMA20/SMA50 on AAPL

A Qiita tutorial shows how to pull Apple (AAPL) price data with yfinance, generate buy and sell signals from S…

Qiita 機械学習Oct 3, 2026

Auto-memory audit of 41 files finds MEMORY.md orphan

A solo founder audited his Claude Code Auto-memory store of 41 memory files

Zenn AI/MLOct 3, 2026

Grab unifies over 500 agent services on LLM-Kit

Grab published how it moved over 500 internal agent services onto one in-house framework, LLM-Kit, cutting pro…

Zenn AI/MLOct 3, 2026

Same model, different tools: gemini-3.5-flash-lite harness drives AI coding success

Running the shared model gemini-3.5-flash-lite repeatedly on the same Python shopping-cart refactoring task, t…

Zenn AI/MLOct 3, 2026

Karpathy: Software 3.0 arrives as agents flop

At Y Combinator's AI Startup School, Andrej Karpathy said software has changed twice in 70 years: Software 1.0…

Zenn AI/MLOct 3, 2026

Developer's CLAUDE.md shrank 99 to 52 lines twice, unprompted

A developer's CLAUDE.md instruction file grew from 85 to 99 lines, then fell to 52 in one commit, rose to 119…

Zenn AI/MLOct 3, 2026

Cursor's pstack: free guide on agent dev, 2,500 PRs a month

A free guide explains how Cursor's pstack built an environment where AI agents handle development, drawing on…

Zenn AI/MLOct 3, 2026

CLAUDE.md vs auto memory: Claude Code's two memories, split by who writes them

The article lays out the split in Claude Code — CLAUDE.md is the file the user writes with instructions and ru…

Zenn AI/MLOct 3, 2026

Airbnb CTO Ahmad Al-Dahle: AI resolves half of support tickets

Airbnb CTO Ahmad Al-Dahle said about half of support tickets are now resolved purely by AI, 60% of code is AI-…

Latent SpaceOct 2, 2026

Gwen Davis: Direct AI agents, don't just use them

GitHub published career advice from Gwen Davis saying developers must direct AI agents, evaluate their output…

GitHub Blog (AI)Oct 2, 2026

AWS fine-tunes Qwen3.6-27B search agent on Amazon SageMaker AI

AWS fine-tuned a Qwen3.6-27B search agent using multi-turn reinforcement learning on Amazon SageMaker AI, with…

Amazon AI BlogOct 2, 2026Also reported by Top Companies AI

Anthropic Message Batches API cuts LLM token costs 50% off

Anthropic's Message Batches API offers a 50% off rate, takes up to 10,000 requests per batch, and returns resu…

Zenn AI/MLOct 2, 2026

Claude Opus 5 / Sonnet 5 / Fable 5.1 migration guide warns of silent API breaks

A Zenn guide documents moving Opus 4.x / Sonnet 4.x code to Claude Opus 5 / Sonnet 5 / Fable 5.1, flagging new…

Zenn AI/MLOct 2, 2026

Tiny LLM hits 573,952 params, repeat-mumbles at 18000字

A hobbyist built a fully from-scratch tiny language model with 573,952 parameters, a 16-token context, and byt…

Zenn AI/MLOct 2, 2026

Claude Opus 5: code switch alone breaks on 7 points

A migration writeup lists seven breakages for code moved to claude-opus-5 or claude-sonnet-5, all in the offic…

Zenn AI/MLOct 2, 2026

Pi 1.0: Earendil ships MCP and virtual models

Earendil released Pi 1.0 on October 1, 2026, with standard MCP support, Codemode, Deferred tool loading, virtu…

GIGAZINE AIOct 2, 2026

Pi 1.0 and Pi Durable hit HN front page

Earendil's Pi 1.0 and Pi Durable both hit the HN front page

Latent SpaceOct 2, 2026

Claude Code opens to TypeScript mods

Claude Code now lets users change its behavior, customize the UI, and swap in their own features using a few l…

GIGAZINE AIOct 2, 2026

Today's AI, for Today's work.

Get curated AI news from 100+ sources delivered daily to your inbox. Free to use.

60,000+ monthly readers

Get Started Free

Free · 30 seconds with Google · unsubscribe anytime