This is BOT Expert Advice...

Top News

• New York weighs expert chatbot ban: The state legislature is considering a bill that would prohibit AI chatbots from providing many kinds of expert advice, and open companies up to lawsuits should their models break the rules. Specifically, chatbots are barred from giving “substantive response, information, or advice” that violates any kind of professional licensing laws, a far-ranging definition that could include not just medical and legal outputs, but responses about nursing, psychology, social work, engineering, and more. The bill passed out of committee this week by a vote of 6-0.

• Netflix bought Ben Affleck’s AI startup: The streamer scooped up InterPositive, an AI filmmaking company founded by Affleck in 2022. The company developed an AI model that helps human teams make faster and more efficient edits during post-production, allowing them to adjust lighting elements, fix continuity issues and more on the fly. When he announced the project 4 years ago, Affleck said the goal was to “preserve what makes human storytelling human, which is judgment.” While breathless AI social media posts favor the “Hollywood is DONE!” shock value approach, this likely aligns closer with reality: AI tools that keep (maybe fewer) human creatives in the loop, while making them more productive.

• Roblox using AI to censor chats: The gaming-slash-social platform beloved of Gens Z and Alpha introduced an innovative new feature: AI-powered “rephrasing” of filtered or banned language in chat rooms. Currently, if a moderator-bot spots a red flag phrase, it replaces the text with # signs. This new protocol goes a good deal further, actually altering the offensive phrase into “respectful language” while keeping the original poster’s “intent.” (The provided example: If a user posts “Hurry TF up!” the bot substitutes “Hurry up!”) On top of its now-mandatory age verification protocol, Roblox is working overtime to assure parents that it’s still a safe place for their kids.

TWiST 500

OpenAI (yes, them again!) rolled out two brand new models on Thursday, GPT 5.4 Thinking and GPT 5.4 Pro. The first is a reasoning model aimed at complex work, like coding and supervising AI agents. OpenAI specifically noted that GPT 5.4 Thinking is optimized for running agents with less computing power, meaning lower costs for Claw-pilled users like Jason. Pro, on the other hand, is designed strictly for “maximum performance.”

OpenAI also notes that GPT 5.4 is its most “factual” model to date: 18% less likely to make errors, and 33% less likely to hallucinate, than GPT 5.2. (You should probably still ask Claude or Grok for double-confirmation.) It’s also more adept at handling responses that require citing multiple sources. GPT 5.4 is headed to Codex and the OpenAI API, while the “Thinking” version is en route to ChatGPT.

For those keeping track, it’s only been a few days since OpenAI brought us GPT 5.3 Instant, a smoother and faster model designed for “everyday conversations.” It’s neat to have so many new models dropping so frequently, but AI companies should be on guard against fatigue, lest a new model launch become a non-event. – Lon

A message from Squarespace

Turn your idea into a beautiful website! Go to http://www.squarespace.com/twist for a free trial. When you’re ready to launch, use offer code TWIST to save 10% off your first purchase of a website or domain.

This Week in Startups

E2258: Big Wednesday show. We’ve got Erik Voorhees from privacy-forward Venice AI giving us the behind-the-scenes story of becoming OpenClaw’s default model (temporarily). PLUS Logan Allin of Fin Capital shares how he’s using generative AI to place smarter bets on founders. All that and two demos: George Pickett brings us into the OpenClaw Studio and David Kaufman shows us how Siteline helps keep track of agentic activity.

E2257: Jason welcomes OpenClaw Foundation board member (plus founder and investor) Dave Morin to take us inside the organization, and talk about its goals and where OpenClaw should go from here. PLUS diving into the mounting rivalry between OpenAI vs. Anthropic: who does Dave think will win? THEN we’re checking out two more awesome demos from Greg Kara of Runtools and George Yameen of PickleWatch.

E2256: OpenAI raised the largest private funding round in startup history, giving away 15% of the company to Amazon, Nvidia, and SoftBank in exchange for $110 billion. We’re digging into the deal, and why Jason predicts the bendiest J-Curve in history for the LLM industry. PLUS Nick O’Neill checks in from the crypto chaos unfolding in Miami Beach, and THREE more amazing OpenClaw (and related) demos.

TWiST Partner Offers

• LinkedIn Jobs: Hire right, the first time. Post your first job and get $100 off towards your job post at LinkedIn.com/twist.

• Iru: Unify your identity, endpoint security, and compliance into a single platform. TWiST listeners get 20% off when they book a demo at iru.com/twist!

• Deel: Founders ship faster on Deel. Set up payroll for any country in minutes and get back to building. Visit deel.com/twist to learn more.

Our Favorite Tool

If your work depends on conversations — interviews, meetings, calls — you need a Plaud NotePin. You can check it out at Plaud.ai/twist and use code TWIST for 10% off!

SF Live-Work Space Now Available

Need a flexible living and working environment in San Francisco? This thoughtfully designed loft-style residence at 787 Bryant St., the heart of the vibrant SOMA district and the city’s creative hub, is now available for rent or purchase. Check the listing for more details.

The TWiST500 newsletter is the new, updated, and improved TWiST Ticker.