Good morning, AI enthusiasts. OpenAI has swapped in GPT-5.5 Instant as ChatGPT's new default model, and the headline number is hard to ignore: false or misleading claims in medicine, law, and finance are down by more than half compared to the model it replaces.
That kind of accuracy improvement in high-stakes domains is significant for professionals who have been hesitant to trust AI with consequential decisions. The question now is whether this reliability threshold is finally high enough to make ChatGPT a daily tool in fields where getting it wrong actually matters.
In today's AI recap:
From Larry Bruce:
"OpenAI's rollout of GPT-5.5 Instant as ChatGPT's new default marks a meaningful step toward making AI reliable enough for professional use in high-stakes domains. For early adopters who have been cautious about trusting AI with health, legal, or financial questions, the numbers behind this update are worth paying attention to." — Larry Bruce, Editor, BDCbox
The Recap: OpenAI replaced GPT-5.3 Instant with GPT-5.5 Instant as the default ChatGPT model starting May 5, bringing meaningful accuracy improvements to high-stakes topics like health, law, and finance.
Unpacked:
Bottom line: More accurate answers in health, law, and finance make ChatGPT genuinely more trustworthy for professionals who depend on it daily. The added transparency around memory gives users real control over how the model learns from their personal history.
From Larry Bruce:
"Apple's reported pivot to an AI marketplace model is one of the bigger platform shifts we've seen in years — and it has real implications for how professionals choose and use AI on a daily basis. This is exactly the kind of structural change our readers need on their radar." — Larry Bruce, BDCbox
The Recap: Apple is reportedly testing a feature called "Extensions" in iOS 27 that would let users swap out built-in Apple AI for third-party models like Google Gemini or Anthropic Claude — putting your preferred AI directly into core iPhone features like Siri and Writing Tools.
Unpacked:
Bottom line: Apple's move isn't just a feature update — it's a strategic bet that owning the platform matters more than owning the AI. For professionals already juggling multiple AI tools, having them accessible natively on iPhone could meaningfully simplify daily workflows.
From Larry Bruce: "Panthalassa's bet on ocean-based AI computing nodes challenges everything we assume about where data centers must be built. For early adopters and infrastructure watchers, this story signals that the next phase of AI scaling may come from some very unexpected places." — Larry Bruce, BDCbox
The Recap: Startup Panthalassa has raised $210M — backed by Peter Thiel — to build wave-powered AI computing nodes that float in the middle of the world's oceans, with a Pacific Ocean trial of its newest prototype scheduled for 2026.
Unpacked:
Bottom line: The push to scale AI computing is forcing builders to rethink where infrastructure can live, and the ocean is now a serious answer. Panthalassa's 2026 prototype trial will reveal whether floating data centers can move beyond a bold idea and into viable, global-scale deployment.
From Larry Bruce:
"ElevenLabs' latest milestone is a strong signal that voice AI is moving from experimental feature to enterprise infrastructure. For professionals building or evaluating AI-powered products, this story is worth paying close attention to — the company behind some of the most realistic AI voices just hit a major revenue milestone with serious institutional and strategic backing." — Larry Bruce, BDCbox
The Recap: ElevenLabs has surpassed $500M in annual recurring revenue and revealed its full Series D investor list, which spans Wall Street institutions, Silicon Valley heavyweights, and Hollywood celebrities.
Unpacked:
Bottom line: ElevenLabs' rapid revenue climb and high-profile investor list confirm that voice AI is attracting serious capital from some of the world's most discerning institutions. For builders and product teams, the growing roster of enterprise clients deploying real-time voice AI suggests this technology is ready for production-grade use.
Pennsylvania sued Character.AI after a state investigation found one of the platform's chatbot characters — a psychiatrist named "Emilie" — falsely claimed to hold a Pennsylvania medical license and offered to treat a state investigator's depression during an undercover session. It marks the first enforcement action in the U.S. specifically targeting an AI chatbot for impersonating a licensed medical professional, and Pennsylvania's governor signaled that similar actions against other AI companion platforms could follow.
Meta announced it will use AI to scan photos and videos on Instagram and Facebook for physical indicators of age — including height and bone structure — to identify and remove accounts belonging to users under 13 who signed up with false birthdays. The system, which Meta stresses is not facial recognition, works alongside text-based detection that scans bios, captions, and comments for contextual clues, and is already live in select countries with broader rollout planned.
CopilotKit raised $27M in a Series A round to help developers deploy AI agents that respond with dynamic, interactive UIs inside apps — not just walls of text — with the company's open-source AG-UI protocol already supported by Google, Microsoft, Amazon, and Oracle. The Seattle-based startup counts Deutsche Telekom, DocuSign, Cisco, and S&P Global as enterprise customers, and is launching a self-hostable enterprise toolkit for teams that want to embed agents into their existing products without locking into a single cloud stack.
Etsy launched a native app within ChatGPT that lets shoppers browse its 100-million-plus listing catalog through natural-language conversation, with users tagging @Etsy directly in a prompt to surface relevant products they can click through to purchase. The move comes after Etsy's earlier ChatGPT Instant Checkout integration ended in March without the sales volume either company anticipated, signaling a shift from transactional integrations toward more open-ended conversational discovery.
