🧵If you were building this, what would you add?
Thinking about things like:
• shopping list for ingredients
• “random drink” button
• maybe AI recommendations 👀
Built a cocktail app because I never know what to drink 🍸
“Pour Decisions” helps you:
• find cocktails by ingredient
• browse drinks A–Z
• save favorites
• create custom drink lists
Built with Node.js, Express, EJS + CocktailDB API
#buildinpublic#webdev#javascript
You all are overthinking your side projects.
>This guy made digital balls physically bounce off real objects/sticky notes using just a webcam, a projector, and JavaScript.
Go build something fun.
(via ig/bongyunng)
just another insane week in AI, buckle the fuck up:
- Apple banned vibe coding killing apps like $9 billion replit
- an AI cow startup worth $2 billion is controlling cows via a “cowgorithm” (backed by peter thiel)
- google destroyed Figma’s marketcap by 10% after releasing an AI competitor ‘stitch’
- China released a frontier AI model that BUILT ITSELF. automates 30-50% of ai research for their team.
- elon musk / tesla / spacex is building a 1 TERAwatt chip fab + settlement on the moon.
- a father hooked up chatgpt to his daughters piano and created a ‘Guitar Hero’ game (it’s fucking sick)
- Cursor released a new ai model but was exposed for it being a chinese model (kimi)
- speaking of Kimi - they just raised $1 billion. value has 4X’d in 3 MONTHS
- Jensen huang projected $1 Trillion revenue for nvidia through 2027
- anthropic shipped 8 (EIGHT) features THIS week (inc sunday) + openclaw competitor
- anthropic also dominated enterprise ai with a 73% market share of 1st time users
- Jensen huang projected $1 Trillion revenue for nvidia through 2027
- Runway dropped a new video model that *instantly generates video* in milliseconds.
- the Pentagon is starting to train Chatgpt and Grok on classified information for AI-warfare
- nvidia dropped an AI video game graphic model dlss 5 that everyone hated (i think it’s cool fwiw)
> Learn a framework for the interview
> Forget it after the interview
> Get hired
> Join the codebase
> Nothing makes sense
> Google everything
> Ask AI to fix bugs
> Push random commits
> It works
You are now a software engineer.
Built end-to-end:
• Node.js / Express
• MongoDB
• EJS
• Google Vision OCR
• OpenAI
• Cloudinary
Always open to connecting with builders and teams working on document-heavy systems.
Most PDF parsers work great…
until users upload absolute chaos.
So I built Quantio — a full-stack pipeline that turns messy financial documents into structured insights.
PDFs, scans, receipts — all handled.
Stack:
• PDF parsing
• Vision OCR fallback
• OpenAI analysis
Thread👇
Biggest lesson:
Designing for the happy path is easy.
Designing for:
• blurry scans
• skewed receipts
• half-broken PDFs
• inconsistent formats
That’s where things get interesting.