DAVID BLUNDIN

[DAVID BLUNDIN @BlundinDavid]

@CheddarFlow Per capita would make a lot more sense

[Marc Mieher @marcmieher]

Moonshots vs Moonraker • Follow the Compression

For whatever reason, after rewinding a Moonshots episode to review @alexwg's superb soliloquy on the recently discovered J-Space, I could not help thinking of the centrifuge scene in Moonraker.

Moonshots is my favorite show. I anxiously await the twice-weekly podcast. You can see the rest of this show at
https://t.co/ou4vDZhxdn

Thank you @peterdiamandis, @alexwg, @SalimIsmail and @DaveBlundin  for putting on such an educational, informative and optimistic show!

Ironically, no AI was used in this production.

Follow the compression.

#moonshots #moonraker #jspace

[DAVID BLUNDIN @BlundinDavid]

@DavidSacks All of this overlooks the possibility that Dario is overjoyed to not put Fable on the market. The Admin gave him an easy PR way to pull it and blame the government. If David’s and Dario’s comments are genuine (which they always have been) then this must be. Now why?

[Michael R Dawley Jr @mrdj1968]

Takeoff to Singularity: Anthropic's Global Pause, Argentina's AI Gamble,... https://t.co/o3c2RL2lH4 via @YouTube

This Moonshots episode felt less like another AI conversation and more like a warning flare: recursive self-improvement is here, Argentina is offering machines a legal runway, and the labor market is already being pulled into the redesign.

@PeterDiamandis, @DaveBlundin, @alexwg and @salimismail

[DAVID BLUNDIN @BlundinDavid]

@paulg Bos to NYC to DC is a better example. 4-5 hours. Train, plane and car are all viable. When the Tesla can self park after dropping you off that’ll break the tie. Throw in Starlink and you’re self driving every time.

[DAVID BLUNDIN @BlundinDavid]

@elonmusk All of our data-gathering investments are killing it (Mercor, Mecado, ThinkStruct)… if a 10T model can be pretrained every 2 months, humanity can have a new domain-specific super intelligence as quickly as we can find data. We need
medical, subatomic, and materials data asap.

[DAVID BLUNDIN @BlundinDavid]

@JosephPolitano This data is completely wrong. Like wildly wrong. Explain?

[DAVID BLUNDIN @BlundinDavid]

@ebbyamir Delete and reinstall your Grok app if the App Store doesn’t let you update. Otherwise you won’t see it.

[DAVID BLUNDIN @BlundinDavid]

@xanamini @liquidai Implications of this are much bigger than might appear. Basically we’re approaching a massive global shortage of inference time compute (and electricity). As the new foundation models - Grok4 GPT5 and Gemini3 roll out, it’ll be clear that we need much more efficient inference

[DAVID BLUNDIN @BlundinDavid]

@balajis Interesting thesis, cool thoughts. Two questions, first, the CCP wasn’t even aware of DeepSeek when the open sourcing decisions were made, right? It’s a quant hedge fund that isn’t involved in hardware manufacturing. thoughts? Second, Meta is also doing it. Why? Same motive?

[DAVID BLUNDIN @BlundinDavid]

@gregisenberg That all sounds right. Now predict the end state for us. The timeline on 1, 2, and 3 is just a few years. So we need to be building business models for 4 and beyond. Thoughts? I know, it’s very hard to see beyond the light cone. But maybe use a “chaos theory, bounds” pov.

[DAVID BLUNDIN @BlundinDavid]

@WilliamBryk Anything that leads to self-improvement of the core algorithms will lead to AGI. The AIME and CodeForces benchmarks are pretty convincing that this milestone is close. The people saying, "this will not be AGI" are interpreting this too narrowly.

[DAVID BLUNDIN @BlundinDavid]

@The_AI_Investor Eric is correct, of course, about the demand. But Nvidia is fabless and will be constrained by TSMC's ability to manufacture. Before rushing out to buy the stock, have an opinion about the supply-side constraints. It takes a minute to build new fabs. And then there's Apple..

[DAVID BLUNDIN @BlundinDavid]

World Economic Forum Annual Meeting https://t.co/3dpmhs7eGZ #WEF23 via @social_handles.twitter