TypeSafe AI

TypeSafe AI · Launch Video Breakdown: Hook, Pacing & Motion Design

The first public System One Model. Jev gives AI the properties and guarantees of compiled code.

0:00 · The Hook · Founder Authority Drop
0:00 / 0:00
AI Frontier ModelsSeedSeptember 15, 2026

Technical Authority Drop

14

Cuts / min

2:56

Runtime

39.5M

Views

76,076

Likes

8,214

Reposts

2,940

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Founder Authority Drop

    “I'm Diogo Almeida, founder of TypeSafe AI, and at OpenAI I co-created ChatGPT and RLHF.”

    On screen
    DIOGO ALMEIDA founder of TypeSafe AI CO-CREATED ChatGPT RLHF
    Camera
    Front medium shot cutting to profile medium close-up with technical HUD overlays
    Motion
    Vector HUD bracket bounding boxes with line-drawing callouts
  2. Problem Agitation

    Questioning Chat as AGI

    “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?”

    On screen
    RLHF The post-training algorithm behind most frontier AI INSTRUCTION Human Response Instruction Following ChatGPT Chat We asked ourselves Are models that are superhuman at chat AGI? But The trillion-dollar question Why Not?
    Camera
    Dynamic grid tile transitions alternating with founder dialogue shots
    Motion
    Grid matrix tile flip transitions and kinetic typography layout
  3. Problem Agitation

    The Flaws of RLHF and Autoregression

    “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.”

    On screen
    LM berts LLM pre-trained models RLHF chat models That are optimized for human preferences Mode dropping Overconfidence Lack of reliability These flaws mean that LLMs require humans-in-the-loop No true automation can be done. At TypeSafe we've spent 2 years building in stealth We're now ready.. New type of foundation model Automation
    Camera
    Front and profile alternating cuts interspersed with stark graphic typography cards
    Motion
    Dither-pattern background textures, rotating arc typography, and tracking HUD tags
  4. Product Reveal

    Introducing System One Models & RLCD

    “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.”

    On screen
    System One Models New Architecture Sampler Training algorithm RLCD Reinforcement Learning Calibrated Decisions The Improvements are clear
    Camera
    3D wireframe ground plane push-in into expanding branch node graph
    Motion
    Retro vector wireframe floor with animated orthogonal flowchart branching
  5. Feature Teaser

    Sequential vs Parallel Computation

    “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.”

    On screen
    TYPESAFE LLM Look at how the LLM generates sequentially Which is great for a natural conversation But useless for computers LLMs Extract intelligence from Autoregression Similar to the jump that transformers made over RNNs We are replacing.. Sequential computation Parallel
    Camera
    Side-by-side terminal code capture zooming out to animated conceptual diagrams
    Motion
    Real-time terminal race capture, isometric dot grid fill, and radar-scan wipes
  6. Feature Teaser

    Calibrated Confidence & Zero Error Rate

    “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.”

    On screen
    Confidence level Confidence 0.0 Low 1.0 High Instead of words Structured output error rate TypeSafe Jev-1.0 0.00% Reliable Fast Self-consistent Type-safe Possibilities Applications
    Camera
    3D isometric bar chart perspective view cut to benchmark comparison table overlay
    Motion
    Extruded 3D bar height mapping with volumetric threshold plane and dynamic horizontal bar graphs
  7. Product Reveal

    Releasing Jev: 100x Faster and Cheaper

    “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.”

    On screen
    We're releasing Jev The first Public System One model 94x faster Real-time AI is finally possible 99x Cheaper 100x Cheaper $42 / billion = $0.042 / million Intelligence, at a fraction of the cost. Input Output FREE Jev (TypeSafe AI) Intelligence / Cost Jev This is just the beginning What you build to see with Jev
    Camera
    Hyperspace warp speed lines into horizontal comparison bar charts and scatter plot graph
    Motion
    Radial particle starfield burst, logarithmic scatter plot animation, and founder profile cut
  8. Call to Action

    Motto, Call to Action & Humorous Disclaimer

    “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.)”

    On screen
    We're building Prod Not God typesafe typesafe.ai DISCLAIMER Do not use Jev if you enjoy waiting seconds for responses or paying for words you never asked for...
    Camera
    Isometric isometric logo reveal, split-line framing, settling on pharmaceutical-style disclaimer scroll
    Motion
    Vector logo morph, rapid split-screen typographic lockup, and pharmaceutical speed-read crawl overlay

Related AI Frontier Models Product Launches

Explore all AI Frontier Models launches →
Anthropic
@claudeai
AI Frontier Models
@claudeai
Anthropic@claudeai