Excited to be building my dream AI inference platform. We're scaling quick and developers are loving it, DM me for $100 in free credits.

Parasail and Positron AI partnered to build Snap-Serve, giving customers fast, private, always-on access to 3B and 8B language models at just $30–60/month. Check it out today: positron.ai/snap-serve

Moonshot AI has surpassed xAI in token market share, just a few days after launching Kimi K2 🎁 We also just put up a free endpoint for Kimi - try it now! 👇

Andrej Karpathy Dreaming is the answer!

I think im going to take the rest of the day off and just bask in the glory of this shit post

That’ll solve the rate limit issues

What if Amazon's strategy in the AI cloud market is to just lose?

What do Roblox, Anduril, CrowdStrike and Apple all have in common? None of them looked like obvious winners from the start. But they built something far bigger than anyone expected. We call these companies Modelbusters. Modelbusters either reveal a market that’s much larger

500B tokens served to a customer in a batch job. Total cost: about $2000. Thank god it was all automated. Parasail

GPT-5 thinking seems way more skeptical of its own built in knowledge and info from the internet, based on the thinking summaries. It wants to double check everything. Huge improvement if it actually works

If we’re secretely shilling our companies in the form of public policy, then the cheapest way to lower average housing prices is massive amounts of low cost LLM tokens.

Point solutions don’t last long in startup land. If you don’t control your destiny with the underlying infrastructure or the end customer, you’re middleware and the vice is always tightening.

Idea → Beta: 300 weeks 1 → 1,000 Matics: 32 weeks 1,000 → 2,000 Matics: 13 weeks 🚀We’re just getting started. 🚀

Mike Henry 😂😂😂

If you’re building AI infra stuff can pop like crazy at a moments notice and the best companies are ready to capture that. If you’re not building for scale from day one, you probably picked a boring space to build.

Somebody must be fundraising: DeepInfra is committed to providing high-quality AI model access while maintaining sustainable operations. We're writing to inform you of upcoming price changes for models you've been using. meta-llama/Llama-3.3-70B-Instruct-Turbo Current pricing:

would’ve missed out on Groq, Vercel and many other great companies. Some companies just get there early, but still have a great enduring vision. Investors who can only operate in momentum mode are not gonna last.

(1998) Retail doesn’t are about Linux (Most) Companies don’t care about Linux Only the pundits do Steve Ballmer was right