Welcome, humans.

Anthropic's smallest AI model got a glow-up on Wednesday. Claude Haiku 5.5 costs $0.10 per million input tokens (a token is a small chunk of text, about three-quarters of a word), which VentureBeat noted matches OpenAI's GPT-6 Luna to the penny.

Anthropic says the average job runs about 75% cheaper than on the last Haiku, and up to 90% cheaper on shorter requests. It also halved the price of Sonnet 5.5's cache reads (the fee for reusing text the model already read). Plus, monthly API credits now come with paid plans for building your own apps: $100 for Max 5x, $200 for Max 20x, and up to $500 pooled for Team.

For scale, a million tokens is roughly the first five Harry Potter books. A dime now buys an AI that reads all five. Ollivander charges more for a wand.

Here’s what happened in AI today:

😸 ChatGPT now builds interactive calculators and charts right in your chat
📰 Common Sense Media rated ChatGPT for Teens an unacceptable risk
📰 Microsoft opened Surface Laptop Ultra preorders at $2,599
🍪 Stuut raised $52.5M to automate chasing unpaid invoices
🎓 Two AIs cracked unsolved WWII Enigma messages; steal their method

P.S.: Join us LIVE later today for an awesome conversation on agentic engineering!

TL;DR: Alex Lavaee from Microsoft Research is showing us how to test and verify coding agents while using Atomic to build a 3D Subway Surfers-style game in the background. If you want to see what reliable agentic coding looks like beyond simple demos, join us live later today →

😸 OpenAI's New ChatGPT Builds Mini-Apps Right Inside Your Chat (Free Users Get It Today)

Ask ChatGPT to plan a Sunday roast for friends, and you used to get a wall of text. Now you get a little app with a guest counter that rewrites your shopping list every time the headcount changes.

That's GPT-6 in ChatGPT, OpenAI's newest AI model, rolling out to the 1.2B people who use ChatGPT each week. Its headline feature is Intelligent UI (UI is short for user interface, the buttons, forms, and charts you click). ChatGPT now decides when a chart, form, or calculator beats plain text, then builds it right in the reply.

Here's what happened:

Plus, Pro, Business, and Enterprise users got GPT-6 Sol on Wednesday (Enterprise access depends on workplace admin settings).
Free and Go users get the lighter GPT-6 Luna starting today.
ChatGPT can now start answering while it keeps thinking; on web searches, GPT-6 Instant started answering 44% sooner than GPT-5.6 Instant.
Intelligent UI lives in the Chat tab only; the models behind Work and Codex stayed the same.

How to try it:

Open the Chat tab in ChatGPT.
Describe the tool you want in plain English, e.g., a savings calculator.
Change the numbers or tap the buttons right in the reply.

Why this matters: If you've ever pasted ChatGPT's answer into a spreadsheet to run a "what if," this cuts out the paste. Gemini and Claude have the model write real code to build each interactive answer, while OpenAI describes a library of prebuilt parts (sliders, charts, forms) that ChatGPT snaps together. That probably makes the layouts cleaner and caps what it can build. The risk for you is the math: OpenAI's announcement makes no accuracy claim for the numbers, so treat a retirement calculator like an intern's spreadsheet and check the formula.

Our take: The feature is catching up; the audience is what's new. Gemini and Claude have shipped interactive answers for months, but ChatGPT now turns them on for 1.2B weekly users, free ones included. A slider looks like software, and software looks trustworthy, so a wrong number in a polished calculator may land harder than the same mistake in a paragraph. The open question: will first-time users check the math behind a clean interface, or just trust it?

FROM OUR PARTNERS

Your notepad now speaks 32 languages

Ever finish a call in French, jump into one in English, and find your notes only make sense for one of them?

Granola, the AI notepad for people with back-to-back meetings, now supports 32 languages. You jot down what matters to you, Granola transcribes the meeting from your computer, and when it ends, you get clear summaries with actual next steps.

Think of it as a super-smart notes app that keeps up with your whole team.

Download Granola and try it in your next meeting.
3 months on us with the code THENEURON.

Try Granola.

🎓 AI Skill of the Day: Give AI the Goal, Not the Recipe

Two AI models just cracked World War II Enigma messages that stumped human researchers for nearly 20 years. Their playbook works on your hard problems too.

Enigma was the machine Nazi Germany used to scramble military messages. Per TechCrunch, developer Carter Leffen gave OpenAI's GPT-6 Astra a loose mission: find an unsolved message in an archive and decode it. Astra researched the history, spotted clues, built its own Enigma simulator (a software copy of the machine), and recovered the readable text of a message unsolved since 2005. Meanwhile, Jack Willis used Claude Opus 5 with far more hand-holding, including an officer's known signature as a foothold.

Steal the playbook:

Give a goal, not a recipe. Describe the outcome and let the AI research first.
Let it build its own tools, like a script, simulator, or spreadsheet.
Feed it one fact you're sure of. A known clue narrows the search.
Check against something real. Enigma expert Frode Weierud verified the first result.

Exact prompts weren't published, and the experts say human know-how stayed central.

Goal: [the outcome you want, e.g. figure out why our Q3 numbers don't reconcile].

First, research the background and tell me what you find. Then build whatever tools you need (a script, a spreadsheet, a simulator) to solve it.

One thing I know for sure: [a fact or clue].

Finish by telling me how I can check your answer against something I already know is true.

Want more tips like this? Check out our AI Skill of the Day Digest for October.

Have a specific skill you want to learn? Request it here.

FROM OUR PARTNERS

Some teams never seem to stop moving. They're on Attio, the agentic CRM.

It’s your always-on revenue engine: agents and workflows build pipeline, chase every buying signal, and move deals forward alongside your team.

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Try Attio now

📰 Around the Horn

Dang, this was so much more wholesome than it had any right to be…

Common Sense Media rated ChatGPT for Teens an unacceptable risk after finding parent alerts never fired during an hour of self-harm testing, and urged OpenAI to restrict it to adults.
Microsoft opened Surface Laptop Ultra preorders at $2,599, shipping Oct 16, with a USB-C port that pops off if you trip on the cord.
Nous Research confirmed a $1.5B valuation after a $90M Series B (NVIDIA and Samsung joined in) and launched Hermes for Businesses, which lets companies run private, customized AI agents.
Biohub, the Department of Energy, and NIH committed $1.8B to open biology data so AI can predict how cells respond to treatments; Google DeepMind, Isomorphic Labs, and Meta added $300M.
Crunchbase reported North American startups raised $92B in Q3, down 35% from Q2 (OpenAI and Anthropic had no new mega-rounds) but up 50% from last year, with about two-thirds going to AI.

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

*We gave AI a wallet. Connect TinyFish, and it finishes jobs on real websites. 30% extra on every top-up. Claim 30% Bonus
Moxie lets you use OpenAI, Gemini, and DeepSeek models inside Claude Code (Anthropic's coding assistant) from your Mac menu bar, and it switches accounts for you when you hit a usage limit.
Stuut chases late invoices, matches incoming payments, and settles disputes for big companies, and says customers collected up to 40% more cash (raised $52.5M).
Melius turns plain-English requests into ad campaigns, images, and videos, built by three ex-Ramp engineers who scrapped their first product after six months (raised $25M).
Communicate turns your help center, past tickets, and product docs into a support agent that answers customers instantly and hands off to a person when it's stuck —free sandbox, then paid plans.
Finbar gives you prebuilt company financial models plus tools to dig through filings and earnings calls, so you can skip the copy-paste into spreadsheets —free tier, Pro is $200/month.
Cubicle shows your AI agents as pixel-art workers in a tiny office, walking to a desk when busy and flashing red when something breaks (it plugs into Paperclip, an agent-management app) —free and open source.