co-founder @ stencil

Everywhere
Joined February 2020
Cyber Wednesday
34
Software that improves itself only works if the people overseeing it can keep up and bottlenecks are constantly cleared. Our discovery loops open more than 30 PRs a day. The bottlenecks is us understanding the fix. Getting your head around one ticket properly (what broke, why, and what the fix does) takes one of us about 15-30 minutes of context switching. So now every ticket gets its own trailer. It makes a short narrated video for every ticket: the problem on screen, then the fix.
45
we rebuilt a QA agent around @typesafeai’s Jev at Stencil. a frontier model plans the test. Jev chooses browser actions. vision handles the visual checks. here it signs in, opens an app, triggers the paywall, and checks a spacing fix. 5 checkpoints passed in 28.3 seconds. every checkpoint has a screenshot, every decision has a trace. i can see Jev driving decisions across our software factory: which issues to prioritize, what to test, when to retry, and when a change needs human review.
1
3
187
my experience is the exact opposite. yes, we're working long hours and pushing hard like never before. but we're thinking all day. instead of pressing enter, we invest all our thinking and effort in designing systems that remove human bottleneck, steering and cognitive load. every engineer on the team is a turbo engine that cracks the toughest problems.
I am done with this shit. It is over. The state of engineering right now is horrible. It has been half a month since I started a new role at a big company. Nobody knows anything here. The specs, code, tests, PRDs, tickets, resolution of those tickets, reports, etc., everything is made by Claude Code. Nobody on my team likes this. They are being forced to ship as much as they can. I have heard multiple times from higher management that pushing code is not a bottleneck, so why are we slow? People are working 12 to 13 hours a day just to press enter. Nobody is reading anything. Humans in corporate are doing nothing on their own. Everyone, literally everyone, from an L1 to an L7 engineer here is doing the same thing. Talk to Claude. There is no sense of victory. Nobody is resolving bugs. In reality, nobody is thinking anymore. Everything is done by LLMs. It is so soul-sucking. I would not mind it, to be honest, if we were at least given the time to check out the code and see what is going where. But no, the goal is to just ship. No matter what happens.
1
100
someone i've never met just compared the Stencil with Lovable and Claude and told the internet what she'd actually use. this is how it starts.
57
Dan Leshem retweeted
Replying to @OpenAI
@OpenAI picked the tools small businesses actually run on for their new ChatGPT for small business program. Proud that @honeybook is on it. Great job team!
5
7
1
68
1,094
one of our builders posted her app to her IG audience this week. bedtime batch: drop in your b-roll, get 60 edited reels back. $99/month. 2 days later: 303 comments, people LOVE it. built on stencil. creators are becoming tech founders.
87
We no longer use Claude or Codex desktop. We switched exclusively to building on cloud + sandboxes. Your laptop is the bottleneck
Given some of the results I'm seeing recently, it's pretty clear Codex is a good harness. But it will seem primitive in 2-3 months and we're about to go through another major evolution in how we use AI at the frontier. The next generation of models need more than your laptop.
3
2,306
The art of vibe coding is to build on strong foundation. 99% of magic is powered by tools like tailwind, shadcn, vite, and platforms like Supabase and Cloudflare. It's about picking the right LEGO blocks.
2
5
195
Built the same landing page using 5 different AI models all with the same 1-shot prompt. Prompt and all result in Thread👇
76
265
19
3,249
863,324
GPT 4.1? At Fine, we have a unique way of evaluating LLM quality for coding tasks. It's simple. Users build REAL APPS. If the model is not good enough - this impacts our bottom line $. Sonnet 3.7 is marked safe for now. @alexalbert__ and the @AnthropicAI keep shipping!
2
6
213
20 Things you can do with your extra free time now that you're vibe coding everything: 1. Go to the gym 2. Pick up a screen-free hobby 3. Start a side project 4. Visit your grandma: Seriously. 5. Read a book. 6. Take a cooking class. 7. Explore your city. 8. Plan a weekend getaway. 9. Catch up with friends. 10. Sleep in. 11. Volunteer. 12. Write. 13. Learn a new language: Not a coding language, 14. Try meditation. .. .. 15. Find a new job.
4
159
Build with @cursor_ai, deploy with @FineDotDev. This is becoming a common workflow: 1. Start the project with Fine. Backend, auth and database are automatically setup. 2. Export to GitHub. 3. Make some changes with Cursor. 5. Fine automatically deploys new version.
1
5
141
Dan Leshem retweeted
Vibe coding can be funny, but security is not a joke. Here is my absolute minimalist security checklist. 7 CRITICAL rules. 🔒 Run this checklist before sharing your app:
1
3
1
12
712
Vibe coding can be funny, but security is not a joke. Here is my absolute minimalist security checklist. 7 CRITICAL rules. 🔒 Run this checklist before sharing your app:
1
3
1
12
712
✅ Validate all user inputs: Never trust information coming from outside your system. It's like checking ID at the door - it prevents attackers from sneaking in harmful code through forms or search fields.
1
1
95
Lastly: If your'e not how to implement any of the above security measures, or if it's implemented - ask your AI! For example, you could use the following prompt: "Review the backend & API routes. Are all APIs authenticated?"
3
83