duke/dku alum, building lightweight ATS https://xitter.cf/t.co/bVy7eSOzyh
Tashkent
Joined November 2021
- Tweets672
- Following385
- Followers101
- Likes7.9K
Day 2 of building my app Worklicant:
I saw couple of vacancies where folks were hiring through email, then cold emailed them offering my app
features added:
- bm25 keyword search with @paradedb
- Organizations can invite users through email
- better UI / UX in many places (bye bye default phoenix components)
The image has been AI-enhanced since my desk is a bit messy
🤖 Made with AI
Since posting about my project on @hackernews , I received 8 emails.
7 from listing/ directory websites.
1 from UGC content creator
Is everyone building directories now?
First steps I took to promote my project Worklicant:
- post on HN
- post on reddit groups "build in public", "side projects", also replying there to posts like "what are you building this week"
- Marketers hitting my email to help me rank well in AI chats...
aziz abdullaev retweeted
A lot of shit we are dealing with in the modern computer systems is an adversarial complexity that serves no purpose except making things harder for the benefit of the system's owners. The main problem is that lot of people believe this complexity is necessary.
Imagine if Elon worked on combustion engine sport cars
I have noticed that most of ATSs require yearly subscriptions to have access to them. Why not pay per vacancy?
PageIndex + Jev
The entire RAG industry is about to get cooked.
Researchers developed a new RAG approach that bypasses almost everything traditional RAG depends on.
- No vector DB
- No data embeddings
- No chunking
- No similarity search
It's called PageIndex.
Instead of splitting your documents into chunks and loading them into Pinecone, it creates a tree index that lets the LLM reason through them like a human reading a book.
98.7% on FinanceBench. Outperforms every vector RAG on the leaderboard.
100% free. Open source.
Got an iPad for my 10-yo sis. Downloaded duolingo so that she will learn English. She did 97 day streak.
However, her behavior changed a lot (thanks to short videos from other apps). More impulsive. Goes to bed late. Glued to it.
Had to take the iPad from her.
Insane
介绍比Jev快50倍，在你设备上跑的laya-mlx!
只在你的设备上占用最高1G内存
Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统
我将其移植到MLX,并且做了一些性能优化!
视频中就是这个模型在我的本地M3Max上玩贪吃蛇
这个模型能够以每秒决策60次的速度玩贪吃蛇！
github.com/mizorewww/laya-ml…
aziz abdullaev retweeted
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
The first day of @elixirconf just concluded. Here is the list of new things I learned today (1/n):
- building set-theoretic types with union, intersectoin, disjoint; Singleton, Composite, Recursive, Parametric types
- Nerves has a desktop app; launching hardware kit soon
- hex ecosystem is well-maintained and safe (above industry standard)
- Supervisor + shutdown strategies for sub processes (observer_cli)
- Building Agents in Elixir; Agents are just actors (genserver); NixOS; Agents spinning custom NixOS sandboxes and using it as a tool