The first public System One Model. Jev gives AI the properties and guarantees of compiled code.
The Hook
“I'm Diogo Almeida, founder of TypeSafe AI, and at OpenAI I co-created ChatGPT and RLHF.”
Problem Agitation
“The post-training algorithm behind most frontier AI. Our team trained the first models to be superhuman at instruction following, what we now call chat. And we asked ourselves: are models that are superhuman at chat, AGI? The answer was obviously not. But the trillion-dollar question is: why not?”
Problem Agitation
“RLHF has led to LLMs that are optimized for human preferences, and include issues such as mode dropping, overconfidence, and an overall lack of reliability. These flaws mean that LLMs require humans-in-the-loop and almost no true automation can be done. At TypeSafe, we've spent two years building in stealth, and we're finally ready to share our new type of foundation model that's optimized for automation.”
Product Reveal
“System One models. With a new architecture, new sampler, and new training algorithm: reinforcement learning for calibrated decisions. The improvements are clear if you see them side by side.”
Feature Teaser
“Ask a System One model a ton of structured questions, just like you would an LLM. Get the answers back near instantly. Meanwhile, LLMs take hundreds of times longer to finish responding. Look at how the LLM generates sequentially, which is great for a natural conversation, but totally useless for computers. LLMs extract intelligence from the tiny straw of autoregression. And similar to the jump that transformers made over RNNs, we are replacing sequential computation with parallel, because that is obviously the future.”
Feature Teaser
“Our System One models output decisions with probabilities and confidence instead of words, and they can't hallucinate. They're a lot more like code: reliable, fast, self-consistent, and type-safe. This opens up a whole new world of possibilities and applications.”
Product Reveal
“Today, we're releasing Jev: the first public System One model. It's a hundred times faster, so real-time AI is finally possible. A hundred times cheaper, with input tokens priced at forty-two dollars per billion tokens, and output tokens are free because they're finally too cheap to meter. Jev is supersmart, and its intelligence per dollar is literally off the charts. This is just the beginning. We're excited to see what you build with Jev.”
Call to Action
“As we say at TypeSafe, we're building prod, not god. (Disclaimer voiceover: Do not use Jev if you enjoy waiting seconds for responses or paying for words you never asked for. Side effects may include fewer retries, fewer parsers, fewer broken schemas, and intelligence running inside every software loop... Ask your developer if your industry is ready.)”
Synthesizable in Remotion · key components of the typesafe-frontier-launch template.
Mosaic grid blocks sequentially scale down and fade out with stippled edges to reveal the underlying frame.
Generate a 10x6 matrix of SVG rects; calculate stagger delay per cell based on Manhattan distance from center using Remotion interpolate and spring.
Bounding box with corner reticles and anchor lines tracking subject movement.
Absolute positioned container with border-corner SVGs animated via spring opacity and scaleX/scaleY properties.
Side-by-side terminal interfaces executing commands simultaneously to contrast execution latency.
Two flex-1 containers with monospace log streams; drive left stream completion at frame 15 and right stream at frame 120.
Isometric 3D bar matrix representing multi-class confidence values sliced by a semi-transparent horizontal threshold plane.
Three.js or CSS 3D matrix where div heights animate with useCurrentFrame spring from 0 to target normalized value with greyscale-to-black color interpolation.
Purpose. Establish immediate credibility through OpenAI credentials.
Execution. Crisp talking head framed by warm wood slats with precision HUD lower thirds.
Purpose. Deconstruct existing frontier LLMs as chatbots unsuitable for autonomous software.
Execution. Stark typography, mode dropping animations, and vintage software cards.
Purpose. Prove technical superiority via concrete real-time benchmarks.
Execution. Terminal side-by-side race, speed factor graphics, and cost scatter plot.
Purpose. Subvert Silicon Valley god-complex tropes with practical engineering focus.
Execution. 'Building Prod Not God' slogan followed by a parody pharmaceutical disclaimer crawl.