Dept of defense accountant reviewing expenses on Monday: “Hey does anyone know what, eh, openrouter is???”

lfg

a clear observation after living in coding agents for the last couple months: IDEs are basically over. the shift in my own IDE usage has been dramatic: i used to spend many hours per day living in an IDE, now i spend mayyybe 30 mins/day in an IDE. almost all my "coding" time is spent managing claude instances via the cli. the IDE going away is an almost obvious conclusion to me at this point. but human's driving cli's obviously won't replace IDE's. the current state is just a bridge to what i think and hope is coming next... i think we'll create a completely new form of "professional" software to manage swarms of coding and non-coding agents, but it won't feel anything like the work software we use today. it will be highly visual, on a 2d (or 3d) canvas, more video game than software. it will feel like an RTS game, but upgraded for the ai era. it will have classic RTS elements -- cursor to select nodes, chat, etc -- but deep voice integration will probably drive many/most of the interactions. maybe even a new tactile interface. something like factorio meets the minority report interface. this won't feel like a retro re-creation of the rts games of my youth. it will probably feel wild, chaotic, maybe even scary at first. it definitely won't feel familiar. figuring out how to manage, motivate, and lead swarms of agents is clearly where we're headed. it could be the great software design problem of this era. we're in the liminal space right now, but whatever comes next, it's going to look very different than a document editor built to manually slap code into a file. excited for the future of making software.

“bitcoin isn’t digital gold it’s a levered risk asset” actual gold: lol hold my beer

hell yes, my guy @whatdotcd on dialectic just dropped youtube.com/watch?v=FxeHtzsx…

lol impressively low numbers

Calling the top on the Zuck talent acquisition strategy. On product: meta is too far behind already (no one uses llama outside of meta platform integrations), and I don’t think metas mission is aligned with building AGI, which seems like what most of these people originally signed up for. It’s great that llama is “open source”, but like any company, meta has a mission and priorities, and the massive investments they’re making in AI are ultimately in service of that. Same for google, Microsoft , et al. If it comes down to AGI that might cannibalize meta’s business and a video generation model to improve its core business, which do you think they’ll pick? Ultimately OpenAI and Anthropic are the only large independent labs focused solely on building consumer focused AGI. On culture: generally I don’t think a top dollar talent acquisition strategy works. Airdropping a collection of random newly very rich people into an existing very large company creates weird cultural incentives. Instead of mission maxxing everyone is earn-out maxxing. And the politics of all these new people meshing with each other and the existing staff will be…awkward water cooler convo. The Yankees tried this for decades and it was a colossal failure. I admire the boldness of the strategy, I just don’t think it’ll work, and will probably end up like the metaverse.