Roman Builder

🧵 I make 6 figures as a software engineer. But every day I wake up 1hr earlier and build a Chrome extension to kill the tab bar. Starting today: 30-day public challenge. Every number. Every mistake. Zero filter. Here's what I'm building and why 👇 [1/10]

@marclou @trust_mrr 11 days vs the traditional 6-month lawyer-volley is the real story. trust_mrr basically built stripe for small-saas exits — the friction was always the legal stack, not finding the buyer

✅ Startup Acquisition #92 on @trust_mrr ✅ This AI startup analyzes TikTok videos to predict if they can go viral. It took 11 days to complete the acquisition (including the Escrow process).

@leerob the 'on call at 2am' part is the actual constraint. agents can write 90% of it but you can't triage what you've never built. that intuition only really sticks when you've done it yourself first

You might believe you should spend less time thinking about code because of AI. I strongly disagree! We’re watching this play out live where tons of AI generated code becomes a liability. At the end of the day, an engineer needs to be responsible / on call for code that gets shipped to production. If you don’t understand the system you’re trying to debug, you’re probably going to have a bad time. Yes, AI can help with all of this, if you set up the proper systems. You can have agents triage prod logs, look at errors, etc. You can speed up parts of the investigation, but an engineer needs to make the call. There might be serious customer or financial implications from that change. I expect the trend continue for trimming dependencies, vendoring code so you can modify it directly, preferring simpler systems with fewer abstractions, and spending waaaay more time thinking about system design and code maintenance. I’ve said this before, but it’s a great time to get familiar with CS fundamentals and some of the history behind what great software looks like. Many parts will be different in the coming years as AI progresses, but also a lot more than people realize will stay the same.

x.com/benjaminshafii… the part nobody admits in the craftsmanship debate: the craftsmanship is the distribution. you can ship the most beautiful product on earth and nobody finds it because you spent the year polishing the corners instead of the funnel

@minchoi the bottleneck isn't patch speed, it's downstream coordination. one model fix needs every consumer of that model to redeploy. AI doesn't help with the lawyer-call between teams

Mythos release feels close. But the real story is not the model drop. It's what happens after: 10,000+ high/critical bugs found. Maintainers asking Anthropic to slow down. Finding bugs is now AI-speed. Patching is still human-speed. What breaks first?

@karpathy @shreyansj the 'clean' part is everyone still knowing who's actually deep in what without needing the title to tell them. that's the thing that breaks first as you scale, not the title itself

@shreyansj It’s refreshing to see a company of this size successfully call bs on the whole thing to this extent. One group of MTS on a mission, clean.

Anthropic did not seriously make Karpathy "Member of Technical Staff"

@benjaminshafii @levelsio the craftsmanship vs ship-fast thing is just survivor bias from 2 different sample sizes. pieter ships 20 to find 1 that sticks. craftsmen ship 1 and bet the whole thing on it. both work, different math

. @levelsio pioneered the “my product barely works has no craftsmanship but, hey it sells” I’m not going to argue the opposite. But having used multiple of his products never ever felt a single ounce of joy. And that won’t change. Why are people looking up to him again?

@markgurman screen-less bands + rings + AI coaching is basically apple admitting the watch was a phone strapped to your wrist. the next wearable era looks way less like glass

Power On: Apple needs to give Watch and Health efforts a shakeup as the wearables market pivots to screen-less bands, rings and AI coaching. Behind the scenes bloomberg.com/news/newslette…

@RoundtableSpace leap (sidebar tabs + spaces for chrome). built it bc i missed arc but didn't wanna leave chrome. today: polishing the AI tab sort that groups your tabs into folders w/o you tagging anything

WHAT ARE YOU BUILDING TODAY?

@theo shipping a 2026 release with a jan 2025 cutoff is the part nobody at google will explain. refinements over fresh pretraining is a real choice they made

Wait wtf, they STILL haven't updated pretraining???

@marclou trustmrr quietly became the only leaderboard where the #2 being a 'stealth company' is actually a flex not a cope

New startup just took over TrustMRR's #2 place 🥈 with $747K MRR

@karpathy the 'back to R&D' framing tells you more about where anthropic actually is than any of their official roadmaps could

Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

@awscloud leap (sidebar tabs + folders for chrome). drag the aws cluster into one folder and the count basically goes to 1 lol

You opened the AWS console to do one thing. You now have 47 tabs open. You are:

Railway is down today so Leap is DOWN. 2 clients already written me that they don't want to use Leap anymore. + vibe - I have some clients - vibe - I can't do anything with this situation. Leap is highly available system but if you pay for PRO you expect not see smth like today

you know which space you're in just by the vibe.

2M+ installs and still shipping core features into the free tier is the right move — every release that doesn't paywall the obvious thing buys another year of trust. respect. x.com/AnasChakrounX/…

Recently shipped Turbo Console Log v3.22.0 🚀 A VS Code extension that automates meaningful log messages for JavaScript/TypeScript/PHP developers. • 2M+ installs • 200K+ active developers turboconsolelog.io turboconsolelog.io/articles/relea… Also increasingly interested in connecting with people strong in developer-tool growth, partnerships, or distribution.

@temporalio @OpenAI forking a live agent into a different sandbox mid-run is the killer move — most of the 'multi-agent overhead' i see in builders' workflows is exactly that state copy/move pain. nice ship.

ICYMI: @OpenAI shipped sandbox support in the Agents SDK recently and we built a Temporal extension for it - fully runnable, lives in their official repo. Durable agents, session management, fork a live agent onto a completely different sandbox provider mid-run. How it works: youtube.com/watch?v=A1H8Jd…

@siddqamar_ai embeddingGemma + sqlite for second brain is the right stack — local + fast. open-sourcing the resume builder is the move too. trust compounds slower than installs but it's the real moat. respect.

i’m qamar. love building stuff for myself and people around me. shipped my first chrome extension this month. also built a resume builder skill and open-sourced it (4 stars so far). now building a second brain using sqlite, embeddingGemma, gemma 4, react + ts. building in public!

@ahmadparizaad shipping the first one is the hardest — cws review + mv3 quirks gang up every time. v1.3 already means you're iterating fast. what does WittyWing do?

100% FREE for limited time My first chrome extension is live on Chrome Web Store WittyWing v1.3 chromewebstore.google.com/detail/wittywi…

@DhruvJain08 @awilkinson @gregisenberg @danshipper yes — the visual layer is exactly why i built leap. each agent session gets its own space, sessions visually separated. 'productive vs chaotic' usually isn't fewer tabs, it's which tabs are in view right now.

@awilkinson @gregisenberg @danshipper The tab management point is underrated. When you're running multiple agent workflows in parallel, context switching between Claude Code sessions in terminal gets messy fast. A clean visual layer makes the difference between productive and chaotic.

The Codex Mac app rocks. Visually I find it way easier to manage 10-15 tabs than Claude Code Mac or Terminal. The battle rolls on! I was a hardcore Claude Code user and when @gregisenberg and @danshipper pushed me to try it I was skeptical. Impressed. A few pieces of feedback that would make Codex sing (CC: @sama and @fidjissimo): 1. Not having the AskQuestionTool available in work mode (only plan mode) is a travesty! Being able to quickly reply vs get a wall of 15 text based questions that I have to type answers to totally takes me out of my flow. (I updated my settings so that it always switched to plan mode whenever it needs my input, but many users won't do this). 2. I can't explain it, but something about the way it updates on its activity / loads / visually thinks, makes it feel slower. 3. Giving the sub-agents names (like human names) is actually distracting. I would prefer to be able to infer what the agent is/does based on its name (Legal Whiz, NextJS Master, etc). 4. If you could solve preference/environment syncing across multiple Macs, that would be incredible. Current Git-based solutions are very hacky and cause all sorts of errors. If I change my settings on my Mac Studio, I'd love it if it synced to my MacBook. 5. It seems weird that it can't control its own integrated browser and use it to click around sites (unless I'm missing something?) Great work! Super impressed!