Text is so 2023
Anthropic's Claude Haiku 5.5 costs roughly 75% less to run than Haiku 4.5 and is pitched as a subagent for Opus and Sonnet, while Max and Team plans gain monthly API credits.
AI news for builders and product teamsUpdated Oct 10, 2026, 19:01 UTC
Ben's Bites
Reporting and perspectives from Ben's Bites. Headlines and excerpts link to the original articles.
Latest stories
Newest firstThoughts from San Francisco
Ben's Bites newsletter shares reflections from a week in San Francisco on how different groups use AI, plus headlines including Claude Code mods, Google Docs opening Markdown files, and Instinct joining group chats.
OpenAI DevDay 2026
At OpenAI DevDay 2026, OpenAI launched Dots, always-on personal agents built on a never-ending GPT-6 Astra chat, alongside GPT-6.1 Sol, a Pro 500 plan, Sign in with ChatGPT, Plugins, ChatGPT Space and an Ultrafast mode. Google released benchmark scores for Gemini 4 Argon, which is not yet available to use.
Sonnet 5.5 is worth a try
Anthropic released Claude Sonnet 5.5, which a newsletter contributor describes as benchmark-close to Opus 5.5 in coding with a big jump in understanding images and charts. The same roundup covers Manus 2.0, Cue, GPT-6 Astra, OpenAI agents accessing Australian government websites, and Meta's new enterprise platform.
50 years of tech devices
Ben Tossell built a device timeline website covering 87 devices from 1976 to 2026, with all images generated by Astra, starting in Codex and finished in Factory. The site includes voting for devices users had or wanted, a leaderboard, and is published at bentossell.com/devices.
Back to Claude
Anthropic released Claude Opus 5.5, which it says performs at the level of Claude Fable 5.1 for most tasks and costs 40% less to run than Opus 5. OpenAI released GPT-6 Luna and GPT-6 Sol, both 50% cheaper than their previous versions and available only in Codex and ChatGPT Work for now.
What can you build with Jev
Jev, a classification-style model built to run inside tools, is now open to everyone, and builders are showing demos including sponsor-segment skipping, comment filtering and Gmail search. Meta's personal agent Muse has a Mac app and connector developer platform, though Amazon has begun blocking it from shopping.
New home for Cowork
Ben's Bites reports that Claude Cowork is merging with Claude Chat, with Claude Artifacts getting dedicated Docs and Slides products. The newsletter also covers new Gemini API Live models, the Jev model from TypeSafe AI, the stealth Union Alpha model, Meta One, and Factory's $200M raise.
Good luck slowing this down
Ben's Bites covers Dario Amodei's new essay urging leading AI labs to slow down for safety, with Sam Altman agreeing while Trump and David Sacks reject imposing slowdowns. Sam Altman also said an OpenAI IPO in 2026 would be "ill-timed" given work on alignment, control and safety.
Telling AI to design is hard
Ben Tossell launched Design Words, a work-in-progress site that lets non-designers select pre-filled design styles and components to generate prompt text for AI agents. Writing in Ben's Bites, he said the tool took 38 sessions, 116 prompts and roughly two days of work, and is still open to feedback.
Everyone needs a Muse
Ben's Bites published a newsletter headlined "Everyone needs a Muse" covering ChatGPT Images 2.5, Meta's personal agent Muse, a claimed OpenAI solution to the Navier-Stokes problem, and an Anthropic employee's resignation over safety concerns.
The first GPT-6 model
Astra, described as the start of the GPT-6 family, has been released and is said to top ARC-AGI-3 and Zapier's AutomationBench while costing the same as Fable 5.1. Commentators describe it as powerful but inconsistent, and OpenAI says it has met its automated research intern goal.
Scraping 107M rows of data to build this
Ben's Bites built a site mapping UK council spending after scraping 107 million rows of publicly available payment data, using Codex, subagents, Parquet and DuckDB. The site is still being tweaked before open-sourcing.
Fable 5.1
Anthropic released Claude Fable 5.1, which Ben's Bites describes as faster and easier to talk to, with a new system prompt restricting song lyrics and copyrighted logos. Anthropic also cut the cost of caching inputs for Fable by 75%, making API usage roughly 25% cheaper.
Build your ideas
Anthropic will end Claude Code's 50% extra-usage promotion on September 14, replacing it with a permanent 25% increase to usage limits. The newsletter also reports OpenAI's proposal to cut Cursor's direct model access on November 12, and OpenClaw 2.0's release.
How I built this
Ben Tossell describes rebuilding his personal website with AI coding agents, starting with an agent-app interface prototype in Codex that he scrapped after scope creep, then building a simpler canvas-style site with annotations and a mascot, deployed to bentossell.com.
Who let the agents in
ChatGPT Work has introduced a sign-in flow that pauses at login pages and lets users enter usernames, passwords and 2FA codes in a separate widget rather than in the chat. The credentials are passed directly to the cloud browser, and the signed-in session can be kept for later tasks or cleared in settings.
Agents on your mobile
Anthropic has added remote control updates to Claude Code, letting users start a new session from their phone, with dropped laptop-to-phone connections recovering automatically and faster session loading on iOS. The newsletter also notes Claude Academy, free Anthropic courses and guides for using Claude.
How I built this
Ben's Bites describes how the author set up a fresh personal agent using a folder of instruction and memory files, then revised the plan after testing suggestions from a coding-focused agent. Around 700 readers said they use or want a personal agent, and roughly 50% of poll respondents asked how it works in Claude.
Simple agents
Ben's Bites reports a poll with 996 total votes on personal agent use, and describes personal agents as a folder of files and tools connected to any agentic tool. Headlines include Meta AI's Mac app, Berd by Block, OpenAI pausing model improvements for two weeks, and Ramp launching router.com.