Consumer apps, bootstrapped. I share the playbook.
Joined April 2024
- Tweets4.5K
- Following130
- Followers370
- Likes6K
Real growth alpha below.
Bookmark this, go to replies and sort by likes.
I’m looking for tactical and practical talks like this… please go find me some and put them in the replies with why you thought it was great (like watch it)
Will give up to ten folks $42 each for their best finds — looking for quality folks!
You have to have @XMoney turned on
I have a Claude subscription. Claude Code billed my API key anyway.
If ANTHROPIC_API_KEY is set anywhere in your environment, it can use that instead of your plan.
I found out after a landing page translation job: $100 of API credits spent.
The work was great btw.
Check your env before big jobs.
The easiest way to find creators for your app is doom scrolling.
If you’re looking for creators to promote your app, try this:
When you’re scrolling social media and see a video in your niche that you actually like:
Open the profile.
Find their email or DM.
Reach out right away.
The key is to do it at that particular time, not "save for later".
Your feed already knows your niche because you watch it every day.
This small habit can bring you 3–5 new creators a week.
Your app probably runs Node, or Python, or PHP: critical bugs in all three this week.
That’s the new normal.
The gap between disclosure and attack is hours.
If you don’t have a system for this yet, here’s the simplest one:
- Dependabot: opens fix PRs by itself
- Trivy: daily scan of the built Docker images
- A daily bot that checks OSV + CISA’s exploited-bugs list
Taking the site down is last resort but you at least need to know what's happening to make that decision.
Every Claude update - where did the search bar go this time?
(agents resolving conflicts?)
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
ran out of github actions minutes so I moved CI to my old macbook.
should’ve done it sooner tbh.
I’ve got a spare macbook I’m turning into my AI dev box + little infra server. (will post about that more later)
ran out of github actions minutes, so CI moved there too.
now it’s faster (runs stuff in parallel), it’s free, and it’s on 24/7 anyway so why not.
You think AI agents won’t use your product.
You’re probably wrong.
Agents will be the main way people use software, so we've made our product agent-native:
• MCP server
• Agent skill (SKILL.md)
• Agent sign-in + consent
• Agentic UI - visuals right in the chat
• llms.txt + robots.txt
• A .md copy of every page
• Public pricing + API schema
• WebMCP for browser agents
• Listing the MCP server and skill in MCP directories
BTW there are services that check agent readiness.
Removing our cheapest price increased conversion by 32%.
We’ve been A/B testing one-time purchase paywalls for our app.
Paywall A has 3 options: small / medium / big
Paywall B has 2: medium / big (no cheap option)
A: 38.9% conversion
B: 51.4% conversions
+12.5 pts (+32%) relative lift - fewer choices and a higher minimum price.
If you sell one-time packs, test removing your bottom tier.
the recovery system is quietly bringing people back
We just gave every user another reason to market our app.
A referral program shipped: every user gets a personal link and earns a share of the revenue from everyone who buys through it.
Obviously users can share it with friends.
But the underrated part is that a link posted on a forum or in a guide can keep bringing in users.
If you’re building something, give your users a stake in it.
We got accepted to Claude for Startups program 🥳
- Claude Premium seats
- 1k of API tokens
Thank you @claudeai !
it's time to build a real agentic company.
The singularity is here.
In case you missed what’s happening in AI lately:
1/
AI is doing real math research
OpenAI released 722 math papers written by one AI model.
Over 100 of them tackle problems mathematicians were stuck on for decades.
Many proofs are checked by a computer, so it’s not just the AI saying it’s right.
The wildest one: OpenAI says the same model cracked a $1M prize problem about how water flows. Unsolved for 90 years.
2/
AI beat expert researchers
The hardest part of research is deciding which experiment to run next.
A research group called P-Zero tested this.
Claude Opus 5.5 matched their best human experts with less than half the experiments.
And this skill doubles every 3 months.
3/
AI builds 3D worlds in your browser
Realistic water, landscapes, full games.
Sometimes from one prompt.
This used to take a team of graphics experts.
4/
AI can reverse engineer software
Apps and games ship as code humans can’t read. AI can now turn it back into readable code and rebuild it.
One guy gave it Terraria before bed. He woke up to a working copy he could mod.
People are running Halo in a browser tab.
And it’s not just games. A developer used it to build free copies of 7 Adobe apps, including Photoshop.
Why this matters
The singularity is when AI gets good enough to improve itself.
Each new version builds the next one faster.
Point 2 is exactly that. AI is getting better at building better AI.
Sources below 👇
Sources:
1. AI math research
722 math papers: openai.com/index/sharing-ai-…
$1M water-flow problem: openai.com/index/navier-stok…
2. AI beat expert researchers
xitter.cf/pzeroresearch/status/2…
3. 3D worlds in the browser
xitter.cf/Aurelien_Gz/status/210…
xitter.cf/MengTo/status/21027607…
4. AI reverse engineering software
Terraria, Halo and more: xitter.cf/rehan_shei/status/2106…
Opus 5.5 is next level. I’m genuinely impressed by the three.js details.
I generated a playable boat scene through Japanese landscapes with dynamic weather, day and night lighting, realistic textures and 3D characters.
Live site: valley.mengto.here.now
The water reflections, physics, scenery and architecture are insane. I could look at this all day.
I reverse engineered a competitor’s app with Opus 5.5.
Today I found an app that’s a straight copy of ours.
Then I noticed they’re copying our video format too.
So I downloaded their ios app, routed it through a proxy, and watched every request it made.
Now I know their whole stack, and their algo based on request / response fields.
Will use that to cover our gaps, or maybe something else 💀.
can I improve my old (stupid) robovac in this way too?
Someone on Reddit reverse-engineered their LG TV because the Plex app took 30 seconds to load
Built an entire native Plex client from scratch using Claude Code, Opus, and Fable.
Started in C, moved to Rust, ended up building his own OpenGL UI framework along the way
Same TV. Same hardware. Profile picker now loads in 3 seconds instead of 30. Runs at 60 fps. Supports 4K, Dolby Vision, Atmos. Fully open source
One person. One TV he was annoyed with. AI tools doing the work of an entire engineering team firmware reverse engineering, media playback, GPU profiling, shader optimization, UI architecture, all of it
is it time to sell software stocks? (nfa)
This guy re-created all the adobe apps with opus-5.5, ported them to rust, and opensourced them
something really cool is happening
Readers added context they thought people might want to know
These projects are partial open-source Rust reimplementations of Adobe apps (built rapidly with AI agents), not complete recreations; their READMEs and roadmaps document many missing features and features still in development.
github.com/storytold/phot…
github.com/storytold/vect…
github.com/storytold/vect…
github.com/storytold/film…
github.com/storytold/film…
github.com/storytold/ligh…
getartcraft.com