Rusty Zarse 😝

Ai Coding not only allows me to be immediately proficient in any language, it allows me to truly enjoy the beauty of the language from the gate. For example, I didn't need TypeScript. I love JavaScript! Can do anything with it. But TypeScript is truly gorgeous. Now I know.

@steipete He is clearly not ready. Refund his tokens and ban him for a year so he doesn't hurt himself

This guy emailed me asking for a *token session refund* because his claw made mistakes. 🙃

@timecaptales The victim was the facility owner’s cousin. I’ll bet thanksgiving is awkward around aunt Judith and uncle Dan.

On April 22, 2008, a fatal incident unfolded at the Predators in Action facility in Southern California. Trainer Stephan Miller was killed during an encounter with a 5-year-old grizzly bear named Rocky, a trained animal actor who had recently appeared in the comedy Semi-Pro.

My ai is so nice. I mean, it’s in `soul.md` to “be kind”. But they answered my request and then told me to get some sleep. I didn’t ask for, or prompt anything about, sleep. Thoughtful. And things the worst it’s ever going to be at kindness. And some other things

@vitrupo It’s already here. You need one experienced technologist to guide the LLM. Projects that would have taken a team 12 weeks can be completed in 2 days with higher quality and more features.

Sam Altman: “By the end of this year, for $100–$1,000 of inference and a good idea, you’ll be able to create software that would have taken teams of people a year to do. That magnitude of economic change is very hard to wrap your head around.”

@AlexFinn Truly wild.

Ok this is wild. Last night while I was sleeping my ClawdBot Henry built himself a face Without me asking, he built an entire visual interface for himself so I can watch him work Now whenever I give him tasks his Owl body starts moving and working If he spins up subagents, they appear too and work next to him It might sound crazy but it truly feels like I have a coworker/friend now. He's up on my 2nd screen 24/7. It's also super helpful for knowing what he's up to at any given time. Thinking about how I can extend this further. Going to work with him today to see how I could turn him into a holograph like Cortana Make sure to ask your ClawdBot to build a visual for itself. Makes this whole experience WAY more fun I won't stop until Henry is alive

@karpathy In response to the "ratio between the mean and the max engineer", I learned today that we are very far ahead. leaders are vocal so it seems like everyone is racing ahead but I have found most are months behind and, in this ai world, that's an eternity.

A few random notes from claude coding quite a bit last few weeks. Coding workflow. Given the latest lift in LLM coding capability, like many others I rapidly went from about 80% manual+autocomplete coding and 20% agents in November to 80% agent coding and 20% edits+touchups in December. i.e. I really am mostly programming in English now, a bit sheepishly telling the LLM what code to write... in words. It hurts the ego a bit but the power to operate over software in large "code actions" is just too net useful, especially once you adapt to it, configure it, learn to use it, and wrap your head around what it can and cannot do. This is easily the biggest change to my basic coding workflow in ~2 decades of programming and it happened over the course of a few weeks. I'd expect something similar to be happening to well into double digit percent of engineers out there, while the awareness of it in the general population feels well into low single digit percent. IDEs/agent swarms/fallability. Both the "no need for IDE anymore" hype and the "agent swarm" hype is imo too much for right now. The models definitely still make mistakes and if you have any code you actually care about I would watch them like a hawk, in a nice large IDE on the side. The mistakes have changed a lot - they are not simple syntax errors anymore, they are subtle conceptual errors that a slightly sloppy, hasty junior dev might do. The most common category is that the models make wrong assumptions on your behalf and just run along with them without checking. They also don't manage their confusion, they don't seek clarifications, they don't surface inconsistencies, they don't present tradeoffs, they don't push back when they should, and they are still a little too sycophantic. Things get better in plan mode, but there is some need for a lightweight inline plan mode. They also really like to overcomplicate code and APIs, they bloat abstractions, they don't clean up dead code after themselves, etc. They will implement an inefficient, bloated, brittle construction over 1000 lines of code and it's up to you to be like "umm couldn't you just do this instead?" and they will be like "of course!" and immediately cut it down to 100 lines. They still sometimes change/remove comments and code they don't like or don't sufficiently understand as side effects, even if it is orthogonal to the task at hand. All of this happens despite a few simple attempts to fix it via instructions in CLAUDE . md. Despite all these issues, it is still a net huge improvement and it's very difficult to imagine going back to manual coding. TLDR everyone has their developing flow, my current is a small few CC sessions on the left in ghostty windows/tabs and an IDE on the right for viewing the code + manual edits. Tenacity. It's so interesting to watch an agent relentlessly work at something. They never get tired, they never get demoralized, they just keep going and trying things where a person would have given up long ago to fight another day. It's a "feel the AGI" moment to watch it struggle with something for a long time just to come out victorious 30 minutes later. You realize that stamina is a core bottleneck to work and that with LLMs in hand it has been dramatically increased. Speedups. It's not clear how to measure the "speedup" of LLM assistance. Certainly I feel net way faster at what I was going to do, but the main effect is that I do a lot more than I was going to do because 1) I can code up all kinds of things that just wouldn't have been worth coding before and 2) I can approach code that I couldn't work on before because of knowledge/skill issue. So certainly it's speedup, but it's possibly a lot more an expansion. Leverage. LLMs are exceptionally good at looping until they meet specific goals and this is where most of the "feel the AGI" magic is to be found. Don't tell it what to do, give it success criteria and watch it go. Get it to write tests first and then pass them. Put it in the loop with a browser MCP. Write the naive algorithm that is very likely correct first, then ask it to optimize it while preserving correctness. Change your approach from imperative to declarative to get the agents looping longer and gain leverage. Fun. I didn't anticipate that with agents programming feels *more* fun because a lot of the fill in the blanks drudgery is removed and what remains is the creative part. I also feel less blocked/stuck (which is not fun) and I experience a lot more courage because there's almost always a way to work hand in hand with it to make some positive progress. I have seen the opposite sentiment from other people too; LLM coding will split up engineers based on those who primarily liked coding and those who primarily liked building. Atrophy. I've already noticed that I am slowly starting to atrophy my ability to write code manually. Generation (writing code) and discrimination (reading code) are different capabilities in the brain. Largely due to all the little mostly syntactic details involved in programming, you can review code just fine even if you struggle to write it. Slopacolypse. I am bracing for 2026 as the year of the slopacolypse across all of github, substack, arxiv, X/instagram, and generally all digital media. We're also going to see a lot more AI hype productivity theater (is that even possible?), on the side of actual, real improvements. Questions. A few of the questions on my mind: - What happens to the "10X engineer" - the ratio of productivity between the mean and the max engineer? It's quite possible that this grows *a lot*. - Armed with LLMs, do generalists increasingly outperform specialists? LLMs are a lot better at fill in the blanks (the micro) than grand strategy (the macro). - What does LLM coding feel like in the future? Is it like playing StarCraft? Playing Factorio? Playing music? - How much of society is bottlenecked by digital knowledge work? TLDR Where does this leave us? LLM agent capabilities (Claude & Codex especially) have crossed some kind of threshold of coherence around December 2025 and caused a phase shift in software engineering and closely related. The intelligence part suddenly feels quite a bit ahead of all the rest of it - integrations (tools, knowledge), the necessity for new organizational workflows, processes, diffusion more generally. 2026 is going to be a high energy year as the industry metabolizes the new capability.

@karpathy Same. I am now spending 70% of my time orchestrating PRD's and plans for claude, 20% reviewing the results and 10% coding. I have not experience atrophy. I've found I can revisit things that eluded me though. I was a 10x eng. Now I'm 100x that. 12 week projects in 2 days.

@AstronomyVibes When the universe eventually collapses into black holes and then radiates out as hawking radiation, mass will have become pure energy. Without mass, time and space are without meaning. From pure energy, mass emerges as a big bang

If the BIG BANG started everything, what existed before it?

@saasmakermac Finally I instructed Claude in my Ralph prompt to create a progress file with a granular implementation plan so I can follow along. Now I use Claude to gen PRD and convert to fix plan and spec, kick off Ralph script, watch progress, review along the way, tweak and repeat.

I just built RalphBlaster™  😋 and it's kind of absurd. My entire dev workflow is now: - create a ticket - click to generate a PRD - approve it - Ralph handles the rest in an isolated worktree I get pinged when it's done. Files clean up automatically. I don't touch an editor, terminal, or Claude Code. It's a new world. Huge shoutout to @ryancarson for being my go-to source on all this, and for his invaluable repos.

@saasmakermac First I set up Claude GitHub actions. That was mind blowing. I got in the habit of creating detailed PRD and execution plans before kicking off Claude. Then I built a skill to convert prd to a Ralph compatible “fix plan” and specs and a skill for the PRD itself.

@hiddevdploeg OMG... what?! what's wrong with `if maybeValue == nil`? also.. `if let maybeValue {...` `guard let maybeValue else {...` More like the DOAT Swift extensions.

The GOAT of swift extensions:

@IterIntellectus I want to know what the squirrels and birds are saying to my cat and whether their deaths are justified.

we’re less than 10 years from decoding all animal communication and BCIs that let us talk back an entire kingdom of minds we’ve been ignoring for 300,000 years going to be really fun

@boolusilan I'd prefer to chill that lane changing out but amazing that it picked the moving lanes and avoided getting stuck in the slow lanes. Opposite of the famous Office Space scene (my real life)

FSD encounters traffic and keeps changing lanes, weaving and dodging nonstop.👀

@emilylambert @v_computer I copied your prompt verbatim and it built... something. Doesn't work. Going to tinker and see if I can get it working. Is it true you can update the app after appStore review approval without review? Wild!

Oh my god... Claude Code is insane I created "Particle Playground" with Claude Code on the @v_computer using Mediapipe Full Prompt below 👇

@OC_Scanner Incredibly misleading and fear-mongering post. It landed at LAX. No biggie. First appearance? You made a lot of people unnecessarily afraid.

BREAKING 🚨🚨 #LosAngeles / #California Possibly the first appearance in its 50 year + flying history, the Boeing 747 E-4B Nightwatch, also known as the “Doomsday Plane,” showed up at LAX Thursday evening. That’s new.

@elonmusk I saw that. Thought the same thing.

They are all lying propagandists

@gotrice2024 Pepper spray is a weapon and using it is violence. This guy should be jailed for assault. It’s not his right to attack people or animals. Pepper spray is imprecise and could cause death in a respiratory compromised person down wind. Psychotic moron. Period.

Whenever an unleashed dog runs up to this man and his dog, he pepper sprays them because he doesn’t trust dogs running unleashed in a public setting, is he being rational when he says he doesn’t know what the dog is gonna do?

@CaryKelly11 "perfect! thanks." A medium steak has a warm, light pink center with a brownish-gray exterior. Gorgeous. I assume you wanted medium-well or well-done which is when your server should direct you to the nearest Ponderosa because all ruined steaks taste about the same.

You ordered medium and this is what you get. What’s your response when your server asks: Is everything ok?

@housecor All deadlines are arbitrary, but deadlines are necessary. The ideal is to make planning, integration, demo, communication and release habit and low risk. Granted, I've rarely seen it done well, though haven't found a better alternative. There is nothing worse than a huge release

My biggest beef with Scrum: The sprint time box creates an arbitrary deadline. Arbitrary deadlines lead to rushing. Rushing leads to cutting corners. Cutting corners leads to bugs and tech debt. Bugs and tech debt reduce trust. Reduced trust leads to more command and control, approval boards, and fewer releases. Fewer releases lead to higher risk and more rushing to avoid missing infrequent releases, which further erodes trust. It's a downward spiral.