Happening now at @ImaginationinAction: Link co-founder Dave Blundin interviews @OpenAI founder @Sama re: building in the #AI world. With AI accelerating into all human endeavors, how do we build on existing platforms? #LinkVentures #ImaginationinAction

Every one of our portfolio companies that got into the data center stack is creating billionaires out of its founders. Energy, transformers, chip installation, finding land, dealing with state government…it all seems to be working. Our vertical AI use case companies are mixed bag with some big winners and a lot that get stuck in GTM. Why? It’s the biggest constraint today. There’s such a high demand for compute that it’s increasing in price over time which is unprecedented. That trickles through the full supply chain. Meanwhile, creating a category out of thin air or convincing legacy businesses to adopt AI is an arduous, long term battle. In the end, both will win, but the short to medium term returns on the two sides of that coin are starkly different. In the AMA, people always ask how they can participate and never look inside: go to Tennessee, look inside Colossus, and work outward from there.

At Lightwave, we’ve been working for months on the future of inference. We’re not yet ready to unveil what we’ve done, but today we’re sharing some of the thinking behind what we’re working on. The thesis is simple: for the next several years, the world will be compute constrained. Chips, power, and grid infrastructure can’t be built as fast as demand for AI is growing, and everybody, including us, are feeling it firsthand. That makes every hour of compute a meaningful tradeoff. Two members of our founding team, Dave Blundin and Justin Milligan, set out to answer a question at the center of this issue: what should that compute be used for? Their answer is that nearly all of it should go to three uses: making intelligence cheaper, putting it to work for measurable human benefit, and making it trustworthy enough to rely on. The paper explains the rationale behind each one, why they strengthen one another, and what we think the people controlling compute should do about it. This is the first in a series where we’ll share what we believe and how we think about the technical problems in front of us. Read the paper: lightwave.ai/research/what-c… @DaveBlundin @Justin_m34

Since the beginning, I’ve been arguing that freezing AI progress would be the worst thing we could do. That safety is an engineering problem with engineering solutions, not a risk knob we dial up and down. This paper an extension of that argument. If we’re going to accelerate, we should be deliberate about where compute capacity actually goes. Justin Milligan (@Justin_m34) and I landed on three uses that make sense today: making intelligence cheaper, making lives measurably better for the average person, and ensuring AI reliably does what we intend it to. Efficiency, flourishing, and alignment. Everything else is a waste, until we resolve our compute shortage. First of many from @lightwave_ai. lightwave.ai/research/what-c…

RT @Justin_m34: TL;DR I wrote a paper with @DaveBlundin about ethical compute usage: (lightwave.ai/research/what-c…) I'm excited to post for th…

What will it take for AI to move through the world alongside us? Looking forward to exploring that question together at MIT this Friday.

Framed perfectly by @PeterDiamandis: the company selling you tokens is selling you compute. The company selling you results is selling you labor. That shift is worth a 1000x price point, and @sama started figuring it out.

Sam calls this the biggest red pill moment of his life. One of the young geniuses in our lab, Will Thompson from @Harvard, got access to @cerebras' 1200 token per second Astra model. Only @sama and his top engineers have it, because you have to overpay roughly 10x per token. It’s so fast that by the time you finish typing a sentence, everything you asked for is already done. Before you blink. Once you’ve experienced it, you can never go back.

Anyone with an engineering mindset should reject the idea that there’s a big knob labeled “risk” that we dial left or right. @alexwg says this about nuclear power: we could have engineered the risk down to near zero and kept the full benefit. Instead we treated it like a knob, decided we wouldn’t tolerate the risk, and dialed all progress to near zero. It feels like that’s exactly what we’re about to do with AI. It’s an engineering problem with engineering solutions. We cannot let this happen.

In a single day, @OpenAI dropped 722 math solutions. That’s 81% of the most important math results of the past three years, published all at once. The roadmap matters more than the math. There’s so much to gain in the looping and prompt management they figured out on Navier-Stokes and then applied here. That’s the playbook for the material scientists, the physicists, the biologists, everybody.

Spring 2023 (picture) to Fall 2026 today (graph): .5T parameters became 2.5T, and now 2.5 will become 10T in just a few months. That log chart look will flex up a lot as the singularity continues ->

Life doesn’t get better than this

Architect Labs (@architectlabs) unveiled Redwood back in August, the first AI chip designed end-to-end by AI. Zero bugs in its first FPGA build, and a projected 3.4x the performance per watt of @Nvidia's Jetson. Is this an incredible threat to Nvidia? Absolutely. Stretch Nvidia's lifespan by one week and that's $7B in revenue. So they're acquiring and investing in anything that moves. The real question is whether Nvidia buys this and buries it, or @AMD buys it to close the gap. Look out, cousin!

The @OrnnExchange GB300 forward surface is now available to Ornn Data Enterprise customers. Use Ornn to see @Nvidia GB300 pricing across contract terms, built from real cleared transactions from reserved contracts. See values at data.ornn.com

AI won't enter big companies from the top down. It'll come bottom up from the kids. That's how the Mac happened. Corporate offices ran on PCs and Unix boxes. @Apple made Macs available in colleges, they spread to high schools, and eventually students handed them to their parents. Adoption moved up the generations, not down the org chart. AI is going to follow the same path. Nobody in high school today does homework without it. When they enter the workforce, they won't accept writing code or documents the old way. In my experience executives are talking about AI because it's on the cover of the @WSJ, but very few are actually using it every day.

Some of the best-positioned AI companies in the world don't think of themselves as AI companies. Businesses with massive proprietary datasets mostly think about AI as a productivity tool - just like any other business: write code faster, automate workflows. But their data is really what could make them AI-native leaders. Their main problem is momentum. A startup can recruit with founding equity and a clean sheet of paper. A legacy company has far more valuable assets, but struggles to keep top AI talent engaged. Someone is going to crack the formula for turning valuable legacy assets into AI-native companies. When they do, it's going to start a huge tidal wave.

Could AI’s next infrastructure boom happen in space? @alexwg and I spoke with @Starcloud_ CEO @PhilipJohnston about orbital AI data centers, 88,000 satellites and Dyson swarms. Watch Conversations in Action: piped.video/yA3n4M_eo_Q #AI #SpaceTechnology

Sam Altman says the world should tolerate a “few bad things” happening, in the pursuit for universal super intelligence. This isn't actually about safety IMO. It's about winning the PR war. @DarioAmodei took the moral high ground on pausing and testing, then said he's keeping best models locked internally because it's too dangerous for you to have. That opened the door for @sama to differentiate by embracing super intelligence for all.

See Link Ventures take the stage at @imaginationxyz Summit 2026, joining conversations on the future of AI, entrepreneurship, and what’s next for builders and investors. Thank you to our host @johnkwerner! In case you missed it, full playlist here, with new videos coming out daily: piped.video/watch?v=rpEaOHiw…

The foundation lab world operates in complete contrast to the Elonverse. The labs are five to seven brilliant, tight-knit people producing breakthroughs. That's the @AnthropicAI DNA, and it's how China operates too. @elonmusk's DNA is massive infrastructure buildouts like @Tesla, @SpaceX, @spacexai, Colossus. He hates depending on third parties for key components of the stack. It’s unclear whether that was the right strategy for the 2010s, but for the 2020s it may be the only strategy that ultimately wins.

A day exploring the future of longer, healthier lives at TEDxBoston’s Unlocking Longevity. Thank you to our speakers, everyone who joined us, and WHOOP for hosting us in Boston. #TEDxBoston #Longevity