Eli

The thing about superior technology is that you still have to go win mindshare and tear down faulty incentive structures.

Tons of really good speculative histories on how AI changes the macro-economy. Almost no speculative histories on how AI changes micro like the nature of firms.

Easy to imagine a future where compensation will be tied to a multiple of useful compute budget managed.

Would be fun if programming and knowledge increasingly began to resemble playing Starcraft. Can already see it happening with people commanding armies of Claude Code agents.

It is crazy to me that some still don't see how big our GPU shortage is: - Most context window is <100k - Delayed rollouts of Agents, Codex - Full Sora never released - Veo 3 roll out taking weeks - Claude constant rate limits - Even big clouds default rate limits are…

the real innovation of LLMs is suddenly opening up a few trillion of mainstreet paperwork businesses that were traditionally too small and too weird for private equity, that can suddenly transact at 2/20 on some nebulous ai labor arbitrage trade. never bet against aum growth

The only companies worth building right now are ones where you want foundation models to get 10-100x better.

Neo-education should be solely concerned with matters that a machine cannot grok.

It's gone under the radar how deep the Chinese triads are in the marijuana business domestically around grow operations. It's not hard to find the local news articles (Maryland, Oklahoma, Massachusetts, and far more). Or you can simply talk to your local Asian supermarket…

Is there anyone structuring asset-based financing around token commitments for AI application layer companies?

1x preference sheets for non-leveraged AI-based rollups is a bad deal for founders. Thank you for your attention to this matter.

Almost every important AI company has revisited the assumptions around application layer architecture in the past month.

The quote below, from @WillManidis, is worth the consideration of every venture investor, GP and LP. (It's also core to the mission of @equidam: making idiosyncratic ideas more legible for finance.) As a product of incentives, capital in VC has gravity. It is inclined to…

The AI software monetization (ROI on AI compute) is going to have to wait until 2026 or later. Earlier in the year, the belief was enterprise rollout would be the end of 2025 and it seems from many industry conversations this is being pushed out.

Container-based agent deployment is going to accelerate exponentially over the next 2 years.

i think markdown is going to eat software development

Promising that sauna + plasma donation looks like a viable way to reduce microplastic concentration. Essentially free.

It's a mistake that AI-era education seems fixated on teaching the current curriculum in 2 hours rather than imagining what should be done with the other 5-7. If we aren't careful this is setting up for public schools simply becoming daycare systems for K-12.

AI-2027 is a brilliant piece of speculative forecasting. We need these sorts of forecasts for most other tech. Market maps are unimaginative. One of the reasons the piece works so well: it doesn't work backwards from a speculative future but marches forwards from current…

The reality of drone warfare right now is that there's no good way to protect citizens from insurgency risk and elites from assassinations. The entire stack for communal defense here is going to need to be invented quickly. My guess: SWAT teams are the right place to start…

The biggest threat to SaaS is not that software becomes very cheap, but that certain business processes are largely obviated by a shift to agent labor. E.g. the future looks grim for HR software.