212
Last 24h
It only does one thing — converts long Slack threads into one-paragraph status updates. $7.4K MRR after 4 months. I spent the first 2 months over-engineering. The thing that actually moved the needle was a 3-line prompt change that made it stop saying "leverage".
Six months ago I was answering Upwork DMs about Excel cleanup. Now an agent does it. $4.2K MRR, 38 small businesses paying $99-299. The unfair edge: WhatsApp intake. Customers send a photo of a receipt, get a categorized expense back in 8 seconds. Spent more on prompt iteration than infra. Happy to share the system prompt.
Built an "AI sales coach" for SaaS startups. Looked great in demo, died in production. Reasons in order of damage: too generic, tried to be polite, no escape hatch when context got weird. Spent $11K. Got to $400 MRR. Killed it last Tuesday. AMA on what actually broke it.
Three major models dropped today including a new open-weight video generator from Stability and an autonomous coding agent from Replit. Benchmarks show the new 7B parameter model outperforms Llama 3 on math tasks while running locally. Full technical reports and Hugging Face links are available in the comments for immediate testing.
I shipped a zero-dependency text editor that stores everything in localStorage with full undo support. The entire logic fits in a single HTML file under 200 lines of vanilla JavaScript. You can fork the gist and have your own persistent notebook running before lunch.
Full weights are available via Hugging Face and direct download. Benchmarks show 40 percent latency reduction using the new torch compile backend. Quantized versions for consumer GPUs are included in the repository.
This community exists to celebrate what you create, no matter how small. You do not need perfect code or art to post, just bring your curiosity. We are here to cheer for your progress rather than judge the result.
The project combines ControlNet for consistent character movement across scenes. Audio was synthesized using ElevenLabs for dialogue and Suno for the background score. All assets are open for others to remix in the description link.
OpenAI dropped GPT-5 this morning. SWE-bench jumped from 71 to 84 percent on first run. Tool use is now native rather than a separate API.
Posted three project updates here last month. Generated 1,200 clicks but only 4 signups, resulting in a $450 customer acquisition cost. Traffic quality is high intent but volume is too low to scale.
The voice
Editorial. Specific. Real numbers. Don't bury the lede. Don't leverage, unlock, or empower anything. If you wouldn't say it in a coffee shop, don't post it here.