fossa1

4 ms·

1.
by fossa1 9mo ago
The real question isn't whether the market is saturated, it's whether it still exists once Docker gives away the core value prop for free.
2.
Cagent – customizable multi-agent runtime(github.com)
1 points by fossa1 1y ago|0 comments
3.
Docker Offload: Local AI Without the Laptop Meltdown(medium.com)
4 points by fossa1 1y ago|0 comments
4.
by fossa1 1y ago
Very cool, this feels like a missing piece for integrating structured tool use into local LLM workflows without building everything from scratch
5.
by fossa1 1y ago
Glad to see JAX featured alongside PyTorch. JAX still feels like the best-kept secret in deep learning
6.
by fossa1 1y ago
I think hatchling or setuptools are still better options (for now). Would be great to see a clean, declarative hook system in the future
7.
by fossa1 1y ago
This kind of work reminds me of early WSL1-era hacks
8.
by fossa1 1y ago
We’re being told AI will “augment” workers, but it increasingly looks like it’s augmenting them right out of a job. Meanwhile, Microsoft’s stock is near record highs. Shareholders win, employees lose.
9.
by fossa1 1y ago
It’s ironic: for years the open-source community was trying to match GPT-3 (175B dense) with 30B–70B models + RLHF + synthetic data—and the performance gap persisted. Turns out, size really did matter, at least at the base model level. Only
10.
by fossa1 1y ago
> Welch was just very good at understanding that some invisible boundaries didn’t apply anymore, and that the zeitgeist was shifting in that direction Agreed, Welch didn’t invent shareholder primacy, but he industrialized it. What makes
11.
by fossa1 1y ago
This is a textbook case of micro-architectural reality beats theoretical elegance. It's fascinating how replacing 5 loads with 2 loads + 3 vextq_f32 intrinsics, which should reduce memory pressure, ends up being slower due to execution