Neil Rafer

Founders -- If you want to get grow EVA in this cycle, buy instances not tokens.

Where are the Claw-Era Investors? Have we learned nothing? linkedin.com/pulse/where-cl…

@eldacantu How did you determine this statistic for your clickbait GDL article this morning? Please stop fear mongering. "Most gas stations remained shuttered."

@andrewparker I’m all for paying for instances not tokens (per my pinned tweet) but that avg dev isn’t only using “local” models in a reasonable timeframe.

@rafer Because small models are showing surprisingly strong performance, so local inference will become a thing, and the avg dev's incremental energy bill will be no higher than playing Overwatch.

Imagine what's going to happen when tokens have zero marginal cost

Finally, he admits what the plan was all along. The moon's a massive treasure; Mars was death and defeat. finance.yahoo.com/news/elon-musk…

There is a lot of completely rational talk about an AI Bubble, but it's not inevitable. #llm linkedin.com/posts/rafer_th…

Most leaders view AI guardrails as a set of handcuffs. They’ve got it backward. In the enterprise, guardrails aren’t there to stop people from doing things—they are there to give them the confidence to start. Innovation doesn't happen when you give people unlimited freedom. It happens when you make them feel safe failing. If your team is terrified of breaking a database or hallucinating a legal contract, they won't use the tools. Build the fences, then let the horses run. Great talking this through on the Call Me [Data] podcast - link below #AgenticAI #Governance

@joshk I think transformer puns also belong on this list.

How long until we have the first fund-of-funds dedicated solely to rolling up all the AI rollup funds?? Names: That's How We Rollup Capital? Rollup Your Sleeves Ventures? Rock ’n’ Rollup Fund? newcomer.co/p/inside-the-v…

@chatgpt21 Of course they’re too expensive. OpenAI is subsidizing us insanely to support their valuation. If you get dependent on their tokens, you are not thinking strategically.

GPT-5.1 (Thinking High) is about 300 times cheaper per task than o3-preview (Low) while scoring only a few points lower on ARC-AGI-1. 1 year later intelligence has gotten 300 times cheaper. This is why I can’t stand people who say “wahh the models too expensive” it will become cheaper.

Drumpf is going to try and leverage the China truce ending a week before the mid-term elections.

@aaron_wadler You’re not that much taller than me. It’ll be comfortable for you too. Show up and test it out.

@rafer Okay, okay - you win.

Is there any better feeling than a clean desk?

@aaron_wadler Or standing at the standing desk that I built into my pool. I used it for the first time Monday.

@aaron_wadler No desk

I think about this constantly

What a way to kick off the week! Grateful for customers like this.

@rubenhassid As far as any actual compute scientists can tell, AGI is not possible on semiconductor technologies. I will not see it in my lifetime. Quantum at cloud scale has a pretty good shot at it, but we are 5+ decades from that. Or, we need AI math that is not bound by matrix multiplies.

BREAKING: Apple just proved AI "reasoning" models like Claude, DeepSeek-R1, and o3-mini don't actually reason at all. They just memorize patterns really well. Here's what Apple discovered: (hint: we're not as close to AGI as the hype suggests)

Over 40 years of combined Shopify App Store experience packed into one panel. I don't think this has ever been done before 🤯 See you next week, Toronto!

@VivaAerobus you’re complete inability to manage immigration out of MX is moving all my very freq flights to @flyvolaris

That said, $aapl should run, not walk, to acquire LoveFrom to repeat the NEXT experience by making Ive CEO.

@KyleBarr5 Ive (& the Android XR team that is copying him) is just going to provide us with a new addiction. No effort will be made to "save" us. & AI is slowing down because training these massive models is near impossible. Lack of data is entirely mktg BS to distract investors.