#Geek#OSS#ML#AI#WebDev#Music
I post random stuff here, mostly from what I read. Opinions are my own and not obligatory so you don't need to frown on them
Updated CS 8803 "Large Language Model" class at @GeorgiaTech this year for 2026!
The reading list spans pretraining, MoE, reasoning, RL & self-play, agents, long-context, test-time scaling, diffusion LMs, safety, interpretability, and more.
👉 https://t.co/rzFFz4tgsI
🔗 https://t.co/rD39bWNiOD
Contributing to OSS and opening a PR on GitHub repo was always free.
But now it's more fair to offer "Paid PRs" which allows AI and LLM contribution but although pays for the AI & automation infra that will evaluate and filters those PRs and the reviewers time.
Strange that LLM agents skills has been for almost a year, and no one built an auto-discovery and auto-selection skills index based on the assigned task.
Might try to build it as a tool call.
No one built a phone AgentOS yet?
Where you don't need to look into a screen, you talk to your assistant which has deep integration with the apps and the browser and do the tasks and report back to you.
LLMs companies will make more future money from maintaining, rewriting and pentesting their previous generated SW more than the money that was spent initially to build them.
That was happening in some SW projects, but that cost model will increase
Build something awesome, doesn't matter if you write the code, generate it or hire aliens to code it.
The code was barely the problem from the beginning
Maybe although organize a community events for install/provision fest similar to Linux install fest back in time.
We can start with Qwen3.6 27B then include other open affordable models
Startup idea: Run-it-yourself/Provision-it-yourself service that first teaches you how to host and run different local models and LLMS with premium of renting out GPU machine/cluster for test out your learning + ability to host your own models in the future.
Richard Feynman in his reply to the question "Can Machines Think?" Summarized why building AI is not about imitation of other's behaviors/tasks, but in an engineering sense to achieve some goals in the most efficient way under defined constraints.
They did what they did. Without AI, without computers(thanks to von Neumann we're not programming using punched paper tapes anymore) and mostly without calculators.
Neumann for example enjoyed making complex calculations in his head to sharpen his mind.
The solution for AI and LLMs companies debt, they stop going for students assignments and repetitive job tasks and focus on fully automating building businesses and companies.
Progress is good