ChatGPT 5.6 is now unlimited for free

Welcome back! OpenAI took the message limit off ChatGPT for people who don't pay. A Stanford lab published 16 working viruses that evolution never made. Five companies agreed on a single format for agent plugins, built on a protocol Anthropic wrote. And a 23-year-old chess master walked out of OpenAI to help build a machine that reads thoughts through your skull.

In today's Generative AI Newsletter:

OpenAI made GPT-5.6 Luna unlimited on the Free and Go tiers starting today, so the message counter most people have been working around is gone.

Plus and Pro got some pretty cool updates too.

GPT-5.6 Sol now powers both Instant and deep reasoning on those tiers, which OpenAI says makes answers more factual and more focused.

Sol is the model built for hard reasoning and code. It's also the one Microsoft told its own engineers to default to this week.

OpenAI said unlimited text chats but images and file upload limits still stand.

Nobody has said what happens to response times once every free user stops rationing. You'll find that out today before we do.

In 2019 Khosla Ventures put $50 million into OpenAI, which was then a research lab with no product and no revenue. That stake is worth billions now.

Recently, close to one in every two venture dollars went into AI, and the biggest outcomes went to people who were already inside before anyone agreed it was obvious.

GenAI Works is at that stage now.

We reach 14 million people, work with 300+ AI brands, and run go-to-market for Nvidia, Oracle, Google, IBM and Notion. Revenue grew 2.5x last year, and $2.7 million is already committed.

You can invest from $1,000, with up to 22% bonus shares in Phase 2.

Learn more about the opportunity

In making an investment decision, investors must rely on their own examination of the issuer and the terms of the offering, including the merits and risks involved. Genai Works, Inc. has filed a Form C with the Securities and Exchange Commission in connection with its offering, a copy of which may be obtained here.

Ready to transform artificial intelligence from a buzzword into your personal revenue generator?

HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.

Inside you'll discover:

Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.

Get Your Guide

Stanford's Brian Hie and his lab published 16 working viruses in Science, designed by AI and grown in the lab.

They're bacteriophages, which infect bacteria and nothing else.

The models were never trained on viruses that infect humans, animals or plants, and the team tested them to confirm they couldn't invent one.

The models are called Evo 1 and Evo 2, and they read a mountain of real genomes before writing a single letter.

Then the team pointed them at Phi X 174, a well-studied virus that attacks E. coli, and gave them the one stretch of genetic code every version of it shares. The models wrote the rest. It’s kind of like what Cursor auto-complete used to do with code back in the day.

There are countless ways to arrange those letters and almost none of them produce anything alive. Hie says the patterns the models picked up from evolution are what steer them toward the ones that do.

The team ordered the sequences, put them into E. coli, and got 16 living phages. Hie thinks the approach could make phage therapy a better weapon against antibiotic-resistant infections.

Samuel King, the postdoc who led the work, calls it a proof of concept. There's still no easy way to turn an AI-written genome into a living organism, which is the only reason this reads as a science story instead of a scary one.

Six companies published Agent Plugins, one package format so a single agent extension works across all their products. Vercel started it, not OpenAI.

A plugin is a folder with a small plugin.json file at the root.

Inside go MCP servers, which wire an agent to live tools and data, and Agent Skills, which are reusable sets of instructions.

Build one and it runs in ChatGPT, Codex, Cursor, GitHub Copilot, Kiro and VS Code.

MCP is Anthropic's protocol. The steering committee is Amazon, Cursor, Microsoft, OpenAI and Vercel, and Anthropic is not on it.

The standard is deliberately narrow.

It defines how a plugin is packaged and found and leaves marketplaces, installation, permissions and sandboxing to whoever runs it, which matters because fake Agent Skills already slipped past security scanners earlier this year.

OpenAI timed the announcement for GPT-5's first birthday today. The stranger fact is that Anthropic wrote the piece everything else plugs into and got nothing for it.

Naomi Bashkansky spent 18 months on OpenAI's safety and alignment team. She quit two weeks ago for Conduit, a startup training models to read the human mind without cutting anyone's head open.

She was a chess prodigy first. World School Champion in the girls' under-13, North American Champion in the girls' under-20, and a Woman International Master.

Paid volunteers sit in a San Francisco basement in sensor headsets and talk to an AI for hours while it records their neurons firing. The signal is so faint the team runs off battery packs, because the hum from a wall socket drowns out a thought.

They have about 10,000 hours of recordings. OpenAI's Whisper needed 680,000 hours of audio before it got good at speech.

Bashkansky says that within a couple of years the main way we talk to AI will be with our thoughts.

She puts a thought-decoding headband at 2027, models taking encoded thoughts straight from Conduit's system by 2030, and AI feeling less like a tool and more like a limb by 2035.

She works there, and she thinks a company that can read and write to the brain is worth more than a trillion dollars, so weigh the dates accordingly.

GlideOS reads your spreadsheet instead of your prompt. You upload the sheet your team already runs on, and it treats the data, the formulas and the layout as the spec, turns that into a proper database, then builds an app on top.

Changing something later doesn't mean regenerating everything, since you can highlight a page, a component or a line of text and prompt that piece alone.

Try this yourself:

📸 Follow us on Instagram for fast, visual AI updates in 30 seconds. 

📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.

🐦 Follow us on X for breaking AI news and real-time industry updates.

🧠 Learn how to build your next AI application with practical resources and expert guidance.