$GAS is joining the @Pumpfun hackathon
We’re building the cultural context layer for AI agents.
We’ve already used it to scale to $1m. And now we’re introducing it to the world, starting onchain.
$GAS is now live.
we've been heads down this month, but want to be more transparent about what's going on behind the scenes
this has been the most exciting few weeks for our team by far. the conversations we've had and the people we've met have opened many doors and further strengthened our conviction in what we're building.
that being said, much of what's most exciting to us internally right now - like the people and companies we're working with - can't be revealed publicly until next month, and unfortunately this is out of our hands. we're more excited than ever to be building and will continue to share updates on the business leading up to that announcement.
this truly is just the beginning. there has never been a better time to build.
we've spent this week heads down with some incredible individuals at the forefront of AI - at one of the most consequential companies on the planet
we can't share specifics quite yet, but we're building a rocketship.
so much more to come
we've broken down generative video into a science. AI humans are hard to get right - most generations look uncanny and unrealistic.
it typically takes a human operator hours to land on the right video.
what we've discovered is that multi-agent systems are surprisingly good at this work. when armed with our massive corpus of real human footage, our analysis agents are able to generate dozens of ad variants at once - then identify and give feedback to our image generation agents in a loop.
we create a high volume of UGC ads, and an agent ensures we only launch the ones that look and feel human. given a few rounds of iteration and leveraging early signal (like hook rate) from ad networks, this system is terrifyingly good at generating UGC video ads
we just went to market for dozens of business categories - all at once.
our agents identified, researched, and created ads for each niche, then launched full funnels tailored to each one.
why do this?
deploying horizontally makes our system resilient. when an agent learns to scale one mechanic shop, it learns to scale all of them - real business integration is the quickest path to training our agents in real-world environments.
the machine is scaling the machine. in the coming weeks we will begin to onboard dozens and then hundreds of businesses across the country. the compounding starts here.
OpenAI shuttering Sora = clear signal that our thesis is proving true
Frontier labs are throwing in the towel on end-to-end AI content. Sora operated in a vacuum with no inputs from real social feeds. Endless video content, completely free, and still nobody used it.
It doesn't matter how good the models get - slop is slop
Without a cultural layer, AI video is dead in the water.
our mission is to enable any business to scale their advertising, from local shops to growth-scale startups to massive enterprises.
we talk to business owners across this spectrum every day. the problem is consistent regardless of size: founders and operators know they need to be doing more, but lack the infrastructure or know-how to start. and everyone asks to use what we're building.
the opportunity here gets clearer every day. we have more conviction than ever that what we're building is unicorn scale.
prios going into weekend
- finalize new contract (consumer unicorn in NY)
- onboard another fulltime hire (3 interviews lined up on mon)
- launch new funnel
we're taking a massive swing this month - more soon.
understanding, mapping, and exploiting the semantic space of high performing ads across business categories is the cheat code for seeding any ad account.
our corpus of video advertising data across organic and paid channels is growing. here’s a peek into a tiny slice of our DB - content we’ve synthesized, clustered into similar nodes and connected by semantic edges.
the future of advertising is here, it’s just not evenly distributed.
@yukinthecut@PangasB@Niksss16s this isn't true. they rate limit weird behavior (like agents with no guardrails making 1000 API calls) but they've always done this
competently built systems don't get flagged
gas is rewriting the playbook on the $1T advertising industry. for those who missed it, here's a recap of who we are and what we've done so far:
> @cj_enr is out of stealth as cofounder 2 (ex apple, ramp) alongside @cade_cx
> developer tokens locked for 3 more months
> gas already powers a $1m run rate business
> hiring, onboarding new businesses, and scaling every week
Catch both Gas cofounders live on spaces tomorrow at 10am ET
We're relocking 25% of token supply, answering questions, and giving updates on the business.