Papers with Code
ML5.js
arXiv
Spell
Lobe
ML Showcase
Apple Machine Learning Journal
Amazon Machine Learning
ThreadRecap
ThreadRecap turns WhatsApp exports into structured, timestamped reports you can act on or hand to someone else.
What it does
Who it is for
Solo professionals in real estate, HR, recruiting, law and client services who need a defensible record of what was agreed, plus anyone catching up on long work, family or community threads.
How it works
Export the chat from WhatsApp, upload the .zip or .txt, choose an analysis goal, get the report in seconds. No setup, no integration, no account for the other participants.
Pricing
Credit packs with no subscription required, plus an optional monthly plan for regular use. Free credits on signup.
Languages
English, Portuguese, Spanish, German, Italian and French.
Independent and bootstrapped, launched in January 2026.
Papers with Code ThreadRecapNo ThreadRecap videos yet. You could help us improve this page by suggesting one.
ThreadRecap's answer:
Most summarizers expect you to paste text. ThreadRecap takes the raw WhatsApp export, the .zip with thousands of messages and voice notes, and turns it into a timestamped record you can hand to someone else. Voice messages are transcribed and merged into the same timeline as the text, so a commitment made in a 4 minute audio at 11pm sits in date order next to everything else. The output is not a paragraph of prose. It is decisions, agreements, deadlines, responsibilities and quoted messages, in chronological order.
ThreadRecap's answer:
Generic AI assistants can summarize a chat you paste, but they hit token limits on long threads, ignore voice notes, and give you a different structure every time you ask. Chat statistics tools give you word clouds and message counts, which is not what you need when you are trying to prove what was agreed. ThreadRecap is built for the specific job of turning a full export into a structured, repeatable record. It works in six languages, needs no integration or account for the other participants, and you pay per analysis instead of subscribing.
ThreadRecap's answer:
Solo professionals and individuals who need to establish what was actually agreed in a conversation. In practice: real estate agents, recruiters and HR, small business owners, freelancers dealing with clients, and people documenting a personal dispute over a rental, a partnership or a family matter. A second, lighter audience uses it to catch up on long work or family group chats. Buyers are spread across more than 20 countries, led by the UK, the US, Italy and Switzerland.
ThreadRecap's answer:
It started as a personal tool. I kept receiving long voice notes and endless work threads, and wanted the main points without listening to everything again. I built the first version over a weekend and published it as a proof of concept, mostly to see whether anyone else cared. People did, but not for the reason I expected. Instead of catching up on chats, they were using it to document disputes, prove agreements and produce a record of what was said. That reframed the whole product, and the roadmap has followed the users ever since.
ThreadRecap's answer:
Next.js and React with TypeScript, Tailwind CSS, Node.js on the backend, deployed on Vercel. Voice note transcription uses OpenAI Whisper, analysis runs on large language models, and payments go through Stripe.
Based on our record, Papers with Code seems to be more popular. It has been mentiond 100 times since March 2021. We are tracking product recommendations and mentions on various public social media platforms and blogs. They can help you identify which product is more popular and what people think of it.
Benchmark Primary focus Evaluation metrics System coverage Usability Link HumaneBench AI benchmark Human well being, humane AI principles HumaneScore, flip tests under adversarial instruction, long term well being 15 popular chat models tested across 800 realistic scenarios Designed for chatbot safety research; requires ensemble judging for... - Source: dev.to / 10 months ago
An helpful approach is to browse the state of the art models in paperswithcode. This will give you an idea of the performance of different models on various tasks. - Source: dev.to / almost 2 years ago
I think a way around this would some sort of voting/ popularity system? Papers with code (https://paperswithcode.com/) does this via Github stars sorting. Sure it doesn't mean something is established. But it at least gives some way to filter through the firehose of papers. Love this project btw! I think it has potential (and the timing is right now that everyone is looking for the next "attention is all... - Source: Hacker News / about 2 years ago
Adapting to Evolving Standards: With the rapid progress in deep learning research and applications, staying current with the latest developments is crucial. The checklist underscores the importance of considering established standard architectures and leveraging current state-of-the-art (SOTA) resources, like paperswithcode.com, to guide project decisions. This dynamic approach ensures that projects benefit from... - Source: dev.to / about 2 years ago
Papers With Code is one of the good resources to get you to get started. - Source: dev.to / over 2 years ago
ML5.js - Friendly machine learning for the web
arXiv - arXiv is a free distribution service and an open-access archive for scholarly articles.
Spell - Deep Learning and AI accessible to everyone
Lobe - Visual tool for building custom deep learning models
ML Showcase - A curated collection of machine learning projects
Apple Machine Learning Journal - A blog written by Apple engineers