Editorial illustration for ChatGPT Co-Creator Launches New AI
ChatGPT Co-Creator Launches TypeSafe AI
Diogo Almeida spent years building the research that taught ChatGPT how to hold a conversation. His newest project skips conversation entirely.
Almeida just launched TypeSafe, and its flagship system, Jev, doesn't generate text. It doesn't chat, doesn't write essays, doesn't do any of the things that made his previous work famous. Instead, Jev sits inside software and makes judgment calls, the kind of background decision-making that chatbots handle poorly or not at all.
That's a deliberate bet. Most of the AI industry has spent the last three years racing to make chatbots faster, cheaper, and more fluent. Almeida's read is that a huge chunk of real-world AI work isn't conversation at all.
It's decisions: routing a request, flagging an edge case, deciding what happens next in a piece of software without a human in the loop. Jev is built to handle that layer specifically, and the pitch centers on speed and cost rather than eloquence.
Coming from someone who helped write the playbook on making AI sound human, building a model that never talks is a strange next move. Here's how Almeida explains the thinking behind it.
TypeSafe's Jev is a new type of AI system that lives inside software and makes judgment calls, doing the background decision-making work that chatbots aren’t always the right tool for — at a wild speed and price point.
Why this matters
We keep hearing that the next leap in AI comes from bigger language models with more parameters and more text to chew on. Almeida's move cuts against that story. He helped build the RLHF work that made ChatGPT usable by regular people, and now he's betting on a model that skips text generation entirely. That's a signal worth watching, not because of hype, but because it comes from someone who understood the last paradigm well enough to help create it.
For founders and researchers, the takeaway isn't "text is dead." It's that the people closest to how these systems actually work are the ones most willing to abandon the format everyone else assumes is permanent. If a co-creator of ChatGPT thinks the next useful architecture doesn't output words, that's a reason to question assumptions baked into current product roadmaps. We don't have enough detail yet on what this model does instead.
But the pedigree here means it's not a gimmick to dismiss. Worth tracking closely as more specifics surface.
Related Reading
Browse all OpenAI news →
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
Related Topic OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source
Related Topic OpenAI Calls Apple's Lawsuit 'Rotten to Its Core' in Rebuttal
Policy & Regulation
From Archives Can LLMs Manage Long-Term Physical Tasks Without Human Help?
From Archives Google's Gemini 3.8 Live Tops AI Speech Ranking at USD 0.005/Minute
From Archives OpenAI Funds Biological Data Collection to Aid AI Disease Research
Policy & Regulation
From Archives Musk antitrust suit against OpenAI proceeds after Apple resolves ChatGPT dispute
Business & Startups
Common Questions Answered
What is Jev and how does it differ from ChatGPT in terms of functionality?
Jev is TypeSafe's flagship AI system created by Diogo Almeida that does not generate text or engage in conversation like ChatGPT does. Instead, Jev operates inside software to make background judgment calls and handle decision-making tasks that chatbots typically struggle with or cannot perform effectively.
Why did Diogo Almeida choose to skip text generation entirely with his new TypeSafe project?
Almeida made a deliberate bet that the next advancement in AI should focus on background decision-making rather than larger language models with more parameters. This represents a strategic shift from his previous work on RLHF that made ChatGPT conversational, signaling that specialized AI systems for specific tasks may be more valuable than general-purpose text generation models.
What was Diogo Almeida's previous contribution to ChatGPT's development?
Almeida spent years building the research behind RLHF (Reinforcement Learning from Human Feedback) that taught ChatGPT how to hold conversations and became usable by regular people. His foundational work on conversational AI makes his pivot toward non-text-generating systems particularly significant.
What does Jev's approach suggest about the future direction of AI development?
Jev's design challenges the prevailing narrative that AI advancement depends solely on building bigger language models with more parameters and training data. Coming from someone who helped create the previous AI paradigm, Almeida's focus on specialized decision-making systems suggests that targeted, efficient AI tools for specific software tasks may represent the next meaningful leap in artificial intelligence.
Further Reading
- This $200 Million Startup Wants To Fix AI's Overconfidence Problem - Forbes
- TypeSafe AI Emerges From Stealth With $40M in Funding - Yahoo Finance
- ChatGPT co-creator launched an AI that can't hallucinate: Here's how it works - Digit
- Introducing System One Models & Jev - TypeSafe AI Blog - TypeSafe AI Blog
- Diogo Almeida — AI Engineer Talks - AI Engineer Talks
