A prompt to start from
Paste it into Clueso or any MCP client to recreate this video. The link to it comes with the copy.
Create a 3-minute founder-led launch video announcing a new language model called Jev from a company called TypeSafe AI. Open on a name card for the founder, Diego Almeida. Alternate between three slide treatments: near-black slides with large centered white type, off-white slides with black type, and accent-pink slides for the turning points. Walk through the argument in order: today's models are superhuman at instruction following, so why are humans still in the loop; the flaws that require supervision; a new architecture, sampler and training algorithm; a diagram showing the model generating sequentially one token at a time; the release itself; pricing at 100x cheaper, $42 per billion input tokens; a tongue-in-cheek "side effects" slide; and close on the claim that entire software categories may be rewritten. Keep the typography large and the pacing deliberate, one idea per slide.
How it’s put together
One frame per shot, with what is said over it.
0:02I'm Diego Almeida, founder of TypeSafe AI.
0:13Our team trained the first models to be superhuman at instruction following.
0:24But the trillion dollar question is, why not?
0:39These flaws mean language models require humans in the loop, and almost no true automation can be done.
0:57With a new architecture, a new sampler, and a new training algorithm.
1:18Look at how the model generates sequentially.
1:53Today, we're releasing Jev, the first public System One model.
2:06A hundred times cheaper, with input tokens priced at forty two dollars per billion tokens.
2:32Side effects may include fewer retries, fewer parsers, fewer broken schemas, and intelligence running inside every software loop.
2:55As costs fall, usage may increase dramatically.
Make one like this
Take the prompt into Clueso and change whatever you need.