Simon Collins

@ForrestPKnight They won't build apps from scratch. They'll mix and match from composable agent skills and tool primitives. The platform will provide core skills that can be overridden or extended at the team or user level and skills will be shareable and version-able.

it amazes me how many people think your average joe is just gonna whip up personal software with ai whenever they feel like it. not gonna happen

@ashebytes This is really nice. I'm looking to build something like this for myself this year so it's interesting to compare notes. For me the daily schedule aspect will be critical since I suck at maintaining one. I love your idea of morning/evening rituals.

@MarcJSchmidt Libraries like Prisma and Tailwind are abstractions for humans. AI does not need them. A lot of libraries are like that. There won't be any marketplace for libraries.

All my new code will be closed-source from now on. I've contributed millions of lines of carefully written OSS code over the past decade, spent thousands of hours helping other people. If you want to use my libraries (1M+ downloads/month) in the future, you have to pay. I made good money funneling people through my OSS and being recognized as expert in several fields. This was entirely based on HUMANS knowing and seeing me by USING and INTERACTING with my code. No humans will ever read my docs again when coding agents do it in seconds. Nobody will even know it's me who built it. Look at Tailwind: 75 million downloads/month, more popular than ever, revenue down 80%, docs traffic down 40%, 75% of engineering team laid off. Someone submitted a PR to add LLM-optimized docs and Wathan had to decline - optimizing for agents accelerates his business's death. He's being asked to build the infrastructure for his own obsolescence. Two of the most common OSS business models: - Open Core: Give away the library, sell premium once you reach critical mass (Tailwind UI, Prisma Accelerate, Supabase Cloud...) - Expertise Moat: Be THE expert in your library - consulting gigs, speaking, higher salary Tailwind just proved the first one is dying. Agents bypass the documentation funnel. They don't see your premium tier. Every project relying on docs-to-premium conversion will face the same pressure: Prisma, Drizzle, MikroORM, Strapi, and many more. The core insight: OSS monetization was always about attention. Human eyeballs on your docs, brand, expertise. That attention has literally moved into attention layers. Your docs trained the models that now make visiting you unnecessary. Human attention paid. Artificial attention doesn't. Some OSS will keep going - wealthy devs doing it for fun or education. That's not a system, that's charity. Most popular OSS runs on economic incentives. Destroy them, they stop playing. Why go closed-source? When the monetization funnel is broken, you move payment to the only point that still exists: access. OSS gave away access hoping to monetize attention downstream. Agents broke downstream. Closed-source gates access directly. The final irony: OSS trained the models now killing it. We built our own replacement. My prediction: a new marketplace emerges, built for agents. Want your agent to use Tailwind? Prisma? Pay per access. Libraries become APIs with meters. The old model: free code -> human attention -> monetization. The new model: pay at the gate or your agent doesn't get in.

@badlogicgames We wont be writing software but all the skills we learned from it are still applicable - just at the business orchestration layer - c.f. @Steve_Yegge's Gas Town which draws inspiration from Kubernetes and Temporal. We are just levelling up to a much higher level of abstraction.

I'm sad that the one white colar profession that you can enter without a degree, that allowed people to break socioeconomic barriers (like in my case) might largely be going away, at least in the form of high paying jobs.

@ProtocolAka @p_ferragu So Tesla is building an entire production line and spending millions on it for something that doesn't exist? Got it.

@simoncollins @p_ferragu Sorry to say this but Cybercab doesn’t exist. (Test units not counted)

CES 2026 = The Great Validation Chamber for Tesla. The signal from Vegas is loud and clear: The industry isn't catching up to Tesla; it is actively validating Tesla's strategy... just with a 12-year lag. Two critical takeaways solidify our thesis:

@illyism I would bet that most of that traffic is AI, not human, or at least will be soon. Either way, the cost of UI generation is exponentially trending to zero and AI doesn't need abstractions like Tailwind or even React. All UIs will be AI generated and on the fly. Probably by EOY.

This is wild to watch 1M monthly traffic, $230k/mo SEO value Adam is sitting on a goldmine and doesn't see it I genuinely don't understand

@ProtocolAka @p_ferragu Tesla's FSD is just one of their autonomy moats that others will need to match in order to compete. Which of them have the unboxed manufacturing process used for cybercab or its super efficient wireless charging? $/mile is the metric that counts - none will beat Tesla there.

@p_ferragu NVIDIA’s beta/v1 is as good as Tesla’s FSD v14… don’t see a catch up here in fact Nvidia will likely win this race while Tesla continues its politics and beta testing

@ashpreetbedi This seems quite similar to Hindsight: hindsight.vectorize.io

Poor man's continuous learning: How to make agents better without fine-tuning or retraining. Over the last few months, I've been using a simple pattern that's made my agents noticeably more reliable and useful. It's also been the most fun I've had building in a while.

I remember going on this (called the Gravitron) in Melbourne's Luna Park. The operator sat in the middle controlling it. One time I went on it the operator got a little too cocky and moved around the centre while it was still going. He moved a little too far and the gravity sucked him under the central railings and up against the wall. Fortunately the thing was on a timer and the guy was only slightly bruised and battered with a torn shirt. Lots of fun but not to be messed with.

I have PTSD from these. I remember watching them from the outside as a kid and thinking to myself that it's a guaranteed death trap. I could not understad for the life of me why anyone would willingly go in there. Still don't get it TBH lol

Elon said they would use geofencing and other strategies in Austin to minimise risk. That's not at odds with the "any road" narrative since the technology is ultimately scalable to any road and that's not an example of using Waymo-like HD maps. It's just a risk control strategy for successful rollout.

I agree that Tesla has some strong advantages in the robotaxi race, but I’m cautious about calling them unassailable moats. Tesla’s camera-only approach is indeed unique, aiming to cut costs by ditching lidar and radar. Competitors like Waymo have sensor-heavy systems. Tesla’s vision-only bet could pay off, but it’s a riskier path, especially in edge cases like fog or glare. You’re spot-on about Tesla’s data advantage. With 6.5 million vehicles globally, Tesla collects unmatched real-world driving data daily, which will supercharge its AI training. The barrier to entry is high (but competitors like Cruise use simulation trying to close the gap). Tesla’s vertical integration is a clear strength, letting them design and build the Cybercab for ride-hailing at lower costs. Tesla’s own manufacturing scale - four factories and growing - is another edge. On the FSD chip, I’m skeptical it’s the "most sophisticated". Nvidia’s Orin chips, used by Xpeng and others, match or exceed Tesla’s compute power (250 TOPS vs. 120 TOPS per chip). Chip performance matters, but software matter the most, that's the real bottleneck. A few points you raised need scrutiny. The claim of map-free navigation lacks evidence - Musk recently admitted FSD uses geofencing, contradicting the "any road" narrative. The first-mover advantage also seems off: Waymo’s already serving 200,000 weekly rides, while Tesla’s Austin pilot in June 2025 will start with just tens Model Ys and remote operators. Tesla’s in a strong position with its data, manufacturing, and cost structure, but competitors are currently ahead in deployment and reliability. I’m excited to see this change quickly. Tesla’s moats are very promising.

ROBOTAXI MOATS. Prove me wrong 0. Elon Musk is in charge of the effort. 1. No one else is even attempting to use vision only. Vision only will work as anyone driving a Tesla on FSD today knows without question. And we know it will work in that Elon just said that the test vehicles in Austin have been driving 1000’s of miles without a single intervention. 2. If a company wanted to catch up with Tesla on vision only, the time and expense required would be literally impossible. No one would make that investment with Tesla already capable of deploying a massive fleet. 3. Vertical production. The companies that have the best chance of competing on autonomous ride hail don’t make their own vehicles and aren’t making vehicles that are designed for ride hail. The companies that make vehicles are not even considering ride hail and none have said they are planning to make a dedicated vehicle. 4. Tesla’s inference chip, the FSD chip is already the most sophisticated real world chip on the market by any metric, and hardware five is under development. 5. No company at this time is offering a solution that can be driven on any road on earth with no mapping or experience. Tesla has proven it can even make modifications to the data set for unusual driving conditions by using YouTube video inputs. 6. Tesla’s real world data is unsurpassed and no company has any chance of getting close as Tesla will continue to create more data every day than any other company can create in weeks 7. Tesla’s data center in Texas is already the largest real world end-to-end neural network training computer in the world and Tesla is adding a massive amount of additional compute as we speak. 8. Tesla has 4 factories currently that can manufacture cars at scale, and is the lowest cost producer in each of its markets. Their additional capabilities from AI, and especially from Optimus will give them a massive lead in efficiency that again, because of their own production of Optimus provides an unassailable lead. 9. Because Tesla will be able to scale up to a million or even millions of vehicles by the end of 2026, and because this is not even remotely possible for any other company, they will have a first mover advantage in almost every important market on earth. 10. Tesla has a HUGE fan base and customer base that will prefer a Tesla Ride Hail compared to any other.

@AbleArne @RobGrieves I don't think it requires a completely separate data set because the base model seems quite generalisable in the same way that we humans can switch between sides of the road without too much trouble . It would need some fine tuning though.

@RobGrieves Except for left hand drive? Would that not require a completely separate data set for training?

BREAKING: We have first party confirmation from Tesla Australia Director that there are no regulatory hurdles in Australia for the release of FSD Supervised! "There’s currently no blockers in Australia to releasing the self driving supervised, as we have in North America,”

@iMBitcoinB @serhii_be @shadcn @vladyslavmoroz @evilmartians I'm guessing this is setup for Tailwind4 which uses OKLCH color pallete so if your project was boostrapped with earlier version that uses TW3 with HSB colors then you'll need to convert the tokens.

@serhii_be @shadcn @vladyslavmoroz @evilmartians i tried to update theme in existing app but looks like its not working as expected.

Made small app this weekend with latest @shadcn registry features. Now you can install themes with CLI. - generate themes from color palettes: Tailwind colors, Radix colors from @vladyslavmoroz or awesome Harmony palette from @evilmartians - copy link & install it with @shadcn CLI Link below

@freychu @naval @gregisenberg Awesome, is there any feature in Ahrefs Lite that you think is a super useful feature (for directory building in particular) once you get past a certain level of experience?

@simoncollins @naval @gregisenberg I’m using ahrefs lite! Ahrefs starter wasn’t around back in 2022 but it’s a great option if you’re just starting out 🙂

A year ago I quit my full-time job to build directories. Idk about you but I personally know lots of people who are quitting with zero plan these days. Only a few ppl know that I started plotting my escape on the first week of getting hired. I studied people like @naval and @thejustinwelsh And even created a “business” disguised as a means to talk to 30+ people who turned a side hustle into a full-time business If I lost everything tomorrow, I’d know exactly how to position myself to take another shot at being a full-time builder If you’re curious about everything i did to quit the “smart” way, check out this video I released today youtu.be/vMS4tFj51Ck?si…

I like the aesthetic of bottom right the best but if you're trying to sell memories then I would use the photos in the bottom left option instead, but individually rather than together, in a slowly animating carousel that swipes between them and ends with the photographer as the final image.

If you have to choose one of these heroes, which one will you choose?

Looks nice! The only suggestion I would have would be to run this through a contrast checker such as the accessibility check in Lighthouse to make sure you have enough contrast between the text and background. In particular, the grey text on yellow in the menus probably could do with both making the text darker and making the background lighter or more desaturated.

Designed the new hero section (concept) for SaaS, What do you think about this hero Section, Let me know.

@levelsio LLM + tools + a goal

Cron jobs that call an LLM

100% agree. Soft skills and system design will become even more important. For me it's exciting because of the opportunities it opens up but I can imagine it can be daunting if you're just starting out. At the same time however, the tools for learning quickly have never been better.

"In the next 3 to 6 months, AI is writing 90% of the code, and in 12 months, nearly all code may be generated by AI." ~ Anthropic CEO, Dario Amodei If that’s true… what happens next? → The role of software engineers shifts from writing to reviewing → Developers become AI copilots - guiding, testing, and refining code → Creativity, problem-solving, and system design become the most valuable skills AI won’t replace developers. But developers who leverage AI will replace those who don’t. Exciting? Terrifying? Both? Drop your thoughts below.

@saurabh_irt @excalidraw Excalidraw has this neat feature where it can import mermaidJS diagram markup which most LLMs are pretty good at outputting. Super useful for documenting your projects or for learning. E.g. using Grok.

Using @excalidraw is best way of learning for me using diagrams .. super easy tool .

@charliedbtaylor No worries, look forward to seeing the final design.

@simoncollins thank you for taking the time to do this! i think you’re right it was a bit too dark :)

First wireframe of the new design done! How did I do?