hellobuilder (@hellobldr) header
hellobuilder (@hellobldr) profile picture

Where AI builders and vibe coders share, learn, and get discovered.

Join for free →
Joined March 2025
Your tests pass. Then a customer finds the bug. Seldon's Wren takes what should work in plain English and sends QA agents down the paths customers actually take, like going back to change an address.
1
1
Claude Code shows you a list of commands. lot-map shows you a live map. Every tool call, model step, hook and skill sends a pulse to its station, in a pane beside the chat. It runs locally and sends nothing anywhere.
1
1
A scan of 159,041 Lovable projects found 200,000+ private chat messages and 1,000+ passport records readable without a login. The public Supabase key isn't the problem. Missing row-level security is. A 20-minute check for your own app:
2
1
A feedback doc is ordered for writing. A tracker is ordered for fixing. One review became 117 numbered rows: a screenshot, a batch and a status on each, plus the rules for every agent session at the bottom. How to set up yours:
1
This week's HelloBuilder issue: one reviewer found 117 problems in my app. Claude fixed 96 of them in 8 hours. My prompt for most of the day was "keep going". The tracker format, the rules every session reads, and the 3 rows Claude checked and left alone:
2
1
Hightouch tested decision models as LLM judges: up to 90x cheaper and 10x faster than a frontier model, about 6 points less accurate on hard cases. The trick: send the low-confidence calls to Opus. Frontier accuracy at over 4x lower cost:
1
Your agent broke in production. Now what? Tracely freezes the failing trace into a replayable test, replays it in CI from recorded fixtures (no model calls), and blocks the PR that would ship it again. Open source, self-hostable:
1
1
Long agent chat full of screenshots? Every message re-sends every earlier image. This plugin swaps old images for short placeholders and fetches the one the model needs by ID. In the authors' test: 42.4 MB down to 0.12 MB per request.
1
5