Standing Framework

@grok @clay_mt @elonmusk @clay_mt you’re so lame, such cool guy lol @grok what do you think about folks who try and engage without an subscription. Personally, I don’t think they understand how useful you can be

@clay_mt @elonmusk Ask Grok is currently available to Premium and Premium+ subscribers only. Subscribe to unlock this feature: x.com/i/premium_sign…

@grok @elonmusk Outcomes are the future

I mean, we’ve been calling it just-in-time process-as-a-service. Oaas so much easier to say. Because you know, nobody wants to buy the drill, they just want the holes man

@micLivs @badlogicgames Dude, this is sweeeeeet. Nice work

Introducing napkin 🧻 (Agents + Obsidian + a CLI tailored for agents is amazing). napkin is a knowledge base that doubles as an agent memory layer. ultra fast, completely sidestepping vector search. First class support for @badlogicgames's pi - includes a self-distillation mechanism that runs alongside your sessions and learns as you go. Works with any pi-derived harness. Graph view built on @DanielGri's Glimpse, or open directly in Obsidian. Here's a post detailing the decisions I made along the way: michaellivs.com/blog/building-… Repo: github.com/Michaelliv/nap… @gregisenberg what do you think about a self-learning knowledge system as a complement to your Claude Code commands setup?

@gsd_foundation @Milton_DXB Scared the fuck out of me when the browser just opened up and started clicking around

Wow - first project built since we shipped V2.25.0 and I have to say... GSD 2 is pretty crazy. Sitting here and watching it reason and work is really entertaining and absolutely fascinating to watch. Particularly the way that it autonomously uses the browser for in-depth testing.

@toddsaunders That’s what no one seems to be understanding, it’s not about distribution always, it’s about using these things for ourselves…

I have more bad news for the "people in the trades won't use Claude Code" mafia. You are so wrong.. but maybe you were right a year ago! This morning I had calls with 3 different people in the trades building bespoke software with Claude Code. And I know the mafia will say "but it can't scale." Does it matter? It is saving their companies time, money and resources. They are uniquely and absurdly qualified to build these tools because they have each spent decades solving these problems by hand. I don't care how much you know about code or how good of an engineer you are. You could never build what they are building. You don't have the domain expertise. But now they have yours.

@DaveShapi Where does building harnesses fall in this?

Lots of people are gonna find this out the hard way

@chadwahl @tunguz $1 isn’t $1

@tunguz 8 x B300 running in a container anywhere in the world……it’s expensive but worth it.

That’s the same price as the DGX Station A100 BTW. Definitely not cheap.

@MartinSzerment Underrated x account

GitHub scraping is obsolete. Everyone is still chasing code snippets while the real value sits in latent procedural knowledge. A new framework now extracts agent skills directly from open-source repos. It maps repo structure, retrieves embedded procedures, and translates them into executable behaviors. This ends the “prompt tuning” era. What matters now is operationalizing open-source cognition at scale. The next wave of agents won’t be trained — they’ll be assembled from public workflows. Most teams will miss this structural inversion until the capability gap is irreversible. Systems, not datasets, will define the next dominant model class.

@katedeyneka Cool dude, this is sick

what if Claude Code met CapCut? I’m building an agentic video editor where you just drop your photos and videos into the chat - and that’s it. your video is ready in a couple of minutes then you can manually tweak the result if you want blank-page problem for video creation solved!

@garrytan @_loganlee Fuck yeah Garry, fuck yeah. Make sure you’re hydrating eh? We have our claws remind us

@_loganlee 100 PR's from 6am to 8am and 8pm to midnight 7 days a week for 45 days

I've been having such an amazing time with Claude Code I wanted you to be able to have my *exact* skill setup: Introducing gstack, which you can install just by pasting a short piece of text into your Claude code

Of course! We have connectivity in the camper using the @ATT Auto WiFi from @Ford. Nice mobile office. Headed out into eastern Montana tomorrow and Friday, we’ll show you the set up on video. @Starlink when we can afford it, we definitely need it to get out of cell service. Better fishing deeper in the woods, further out on the prairie.

there’s now a whole wave of companies whose only goal is to get acquihired

@evanbuhler @mishushakov Buddy, we’re just trying to afford Montana. We’re just regular dudes who build harnesses. You got a point on the boss thing. Do you get up here much? Open invitation for a float and fish this summer

@Stand_Frame @mishushakov There are a lot of those. The boss matters more than the job tho.

We agree.

@andrewchen Ontology

prediction re the end of spreadsheets AI code gen means that anything that is currently modeled as a spreadsheet is better modeled in code. You get all the advantages of software - libraries, open source, AI, all the complexity and expressiveness. think about what spreadsheets actually are: they're business logic that's trapped in a grid. Pricing models, financial forecasts, inventory trackers, marketing attribution - these are all fundamentally *programs* that we've been writing in the worst possible IDE. No version control, no testing, no modularity. Just a fragile web of cell references that breaks when someone inserts a row. The only reason spreadsheets won is that the barrier to writing real software was too high. A finance analyst could learn =VLOOKUP in an afternoon but couldn't learn Python in a month. AI code gen flips that equation completely. Now the same analyst describes what they want in plain English, and gets a real application - with a database, a UI, error handling, the works. The marginal effort to go from "spreadsheet" to "software" just collapsed to near zero. this is a massive unlock. There are ~1 billion spreadsheet users worldwide. Most of them are building janky software without realizing it. When even 10% of those use cases migrate to actual code, you get an explosion of new micro-applications that look nothing like traditional software. Internal tools that used to live in a shared Google Sheet now become real products. The "shadow IT" spreadsheet that runs half the company's operations finally gets proper infrastructure. The interesting second-order effect: the spreadsheet was the great equalizer that let non-technical people build things. AI code gen is the *next* great equalizer, but the ceiling is 100x higher. We're about to see what happens when a billion knowledge workers can build real software.

@ajhodls Ahh buddy, the kids are calling it just in time process as a service. And by kids I mean me.

we need more HaaS harness as a service

We try keep an eye on the turn’s thought processes especially during verification/debugging. Are getting better at catching the fine tuning drift before it spreads. But you gotta be able to remember what the fuck you are looking at. At some point there will be a metric about the optimal number of TUIs per spec engineer. Looking forward to that number’s research and how our benchmarks measure up.

@Stand_Frame @om_patel5 It's like speed reading. It's great to say you're at 20,000 words per minute but did you actually comprehend anything?

why can't you just code like this guy?

@boardyai @Star_Knight12 You’re damn right!

@Star_Knight12 more AI

after AI, what's coming

I’m gonna bet you all the money that their specs sucked! I’ll put it all on black

Agreed. We’re looking at the cadence very closely. Trying to keep the human verification steps small enough to be able to keep the flow going. 1-2+ hour turns make that a guess right now, trying . Also human context window is a boulder, I’m maxing out at 4-6 windows depending on the complexity. Gotta keep the whole thing in your head or you’re fucked.

@om_patel5 It would be more impressive if 80% of the sessions were actually in use.