attogram

6 ms·

61.
by attogram 1y ago
Uh.. you mean AGENTS.md ?
62.
by attogram 1y ago
Any guesses who Customer A and Customer B are? Based on the amount of money they spent, I'm guessing Meta is one.
63.
by attogram 1y ago
Toughest bootcamp is a big stretch. Training devs for an AI-first approach is timely and a good thing. As long as it's combined with traditional code knowledge.
64.
by attogram 1y ago
See: https://downdetector.com/status/verizon/
65.
by attogram 1y ago
And there is no reason mcps must use a vector database. It's all just a really enhanced grep.
66.
by attogram 1y ago
Section 7 "Psychiatric insight is glaringly absent from AI safety." seems true. Maybe because psychiatric effects/affects are difficult to measure.
67.
by attogram 1y ago
Or the new theme url, where u can see all the pics and vids: https://www.reddit.com/r/LocalLLaMA/comments/1gkeo6u/an_llmp...
68.
by attogram 1y ago
So now some folks will intentially add in em dashes to get on the leaderboard — oops!
69.
by attogram 1y ago
So summary: Coding agents are smart. Humans are often dumb and do not prompt agents well.
70.
by attogram 1y ago
"Attention Is All You Need" - I've always wondered if the authors of that paper used such a casual and catchy title because they knew it would be groundbreaking and massively cited in the future....
71.
by attogram 1y ago
Will we ever find out if this is for real, or just a disgruntled employee...
72.
by attogram 1y ago
Nice. But kinda specific to using outside LLMs. No mention of using LLMs locally, which could make privacy much easier.
73.
by attogram 1y ago
Can u provide a summary or transcript for those that have not viewed the youtube vid yet?
74.
by attogram 1y ago
Wow that's a deep dive. And wow 2: "The conversation was so disturbing, that I archived it and went on holiday"
75.
by attogram 1y ago
For work: only AI "on the side" used as interactive docs on whatever issue I'm working on. For hobby projects: All-in with extensive use of Google Jules to create/improve my stuff!
76.
by attogram 1y ago
Nostalgic! Now just waiting for the Guru Meditation errors
77.
by attogram 1y ago
It would be interesting to see some benchmarks of this vs, for example, Ollama running localy with no timeout
78.
by attogram 1y ago
"~20 min for the first token" might turn off some people. But it is totally worth it to get such a large context size on puny systems!
79.
by attogram 1y ago
1 hour and 24 minutes of python history, love it!
80.
by attogram 1y ago
Nice that it has a free tier. But how does this compare to Jules/Claude/etc? What makes it different?
81.
by attogram 1y ago
Well that was indeed a deep dive. We need more stuff like this that helps the local-only and privacy-required crowd.
82.
by attogram 1y ago
So is trying to compete with Brave browser?
83.
by attogram 1y ago
Can AI suffer? No. Can humans anthropomorphize AIs, sometimes to an extreme? Yes.
84.
by attogram 1y ago
Super please no - this is AI marketing gone bad. You don't want financial hallucinations messing with your accounts
85.
by attogram 1y ago
What happens after you close the browser? Is the model still stored locally?
86.
by attogram 1y ago
So how is this compared to just running "small" models locally via Ollama or other free inference engines?
87.
by attogram 1y ago
I agree with "Build in Public and Share" - especially the part of making your mistakes and messess public. Learning from the mistakes of others!
88.
by attogram 1y ago
Here is the hugging face page for those that tldr'd the paper: https://huggingface.co/datasets/m-a-p/AetherCode