tiny web tool calculates seven follow-up dates from an appointment and emails them automatically, earning $299 in 28 minutes. Built with under 100 lines of JavaScript, it replaces manual date math and email sending.
Key points
Earned $299 in 28 minutes using the tool.
Implemented with under 100 lines of JavaScript.
Sources
Prompt: make your site readable by AI agents add prompts
Prompt: make your site readable by AI agents add prompts that generate markdown versions of every public page and handle SEO, AEO, GEO and ASO. The prompts are fetched from the links in the post and run in a coding agent. They produce a markdown file per page and expose AI-friendly files.
Key points
Two prompts are provided to fetch and execute in a coding agent.
The prompts create a .md file for every page and make the whole website readable by AI agents, covering SEO, AEO, GEO and ASO.
Sources
- Original post
- Prompt: make your whole site readable by AI agents - Linked page
- scrolllaunch.com - Linked in the post
Material Science Grand Challenges announced with $1M prize
Material Science Grand Challenges announced with a $1M prize pool. Five specific challenges were listed and a prize is offered for solutions.
Key points
Announcement: five Grand Challenges in Material Science.
Funding: $1M prize money for anyone who can solve these challenges.
Sources
OSWorld-Science benchmark introduced for AI agents in scientific research
OSWorld-Science is a new benchmark and evaluation environment for computer-use agents, targeting tasks grounded in real-world scientific research needs. It aims to test whether AI agents can use professional software as effectively as scientists.
Key points
Benchmark: OSWorld-Science provides a suite of tasks reflecting real scientific research requirements.
Goal: Evaluate AI agents' ability to operate professional software in scientific contexts.
Sources
Agents memory 6x growth fixed by load split and hooks
Agents memory grew 6x in 15 days. Fix was splitting loads per session from on-demand reads and moving must-happen rules into hooks.
Key points
Memory grew 6x in 15 days.
Fixed by splitting loads per session from on-demand reads and moving must-happen rules into hooks.
Sources
Amortized Bayesian inference - trains once, fast posterior
Amortized Bayesian inference trains a network once on simulations and returns posteriors in a single forward pass. It contrasts with MCMC where fitting the same model hundreds of times scales cost with every fit.
Key points
Fitting the same model hundreds of times? With MCMC, the cost scales with every fit.
Amortized Bayesian inference trains a network once on simulations, then returns posteriors in a single forward pass.
Sources
- Original post
- dub.sh - Linked in the post
