adam_patarino

4 ms·

1.
Ask HN: What are you using local LLMs for?
1 points by adam_patarino 2mo ago|0 comments
2.
by adam_patarino 3mo ago
When we started work on our logo for Rig I literally told our design agency we cant do butt holes
3.
by adam_patarino 3mo ago
What for? Cost savings? Something else?
4.
by adam_patarino 4mo ago
I always find it amusing when people would rather spend $200 / mo than let their laptop fan turn on.
5.
by adam_patarino 4mo ago
Yes! And we are using it to build Rig AI to make it easier for anyone to do it too! We are post training qwen 3.6 and combining it with a custom inference engine and harness to get the most out of a smaller model.
6.
by adam_patarino 4mo ago
Smaller models can already outperform SOTA and massive models on specific tasks / domains.
7.
by adam_patarino 4mo ago
I like Zed for this. We are considering autocomplete for Rig after we launch our agent. Local would be 1000x faster.
8.
by adam_patarino 5mo ago
Individual experience is rarely an accurate representation of the broader system.
9.
by adam_patarino 5mo ago
Completely agree. It amazes me how some folks think AI is unlike any other technology revolution. History repeats.
10.
by adam_patarino 5mo ago
> Every software developer knows that you can’t make projects go faster just by typing faster. If that were the case we would all be taking typing lessons. So well said. AI is unveiling how the bureaucracy is the slow part.
11.
by adam_patarino 5mo ago
I’ve been talking to a lot of engineers about how they use AI in their day to day and it’s dramatically different than what you see from the hypers. The vast majority use one agent at a time and careful step through code. The main benefit t
12.
by adam_patarino 5mo ago
Gusto makes payroll zero clicks. I’ve used xero and quickbooks and they integrate with many banks and expense management platforms to automate closing.
13.
by adam_patarino 5mo ago
Closing books and running payroll feel like solved problems with today’s saas and high stakes if you mess up. This is one of those areas I would spend more time checking the outputs than it would take me to click the button myself.
14.
by adam_patarino 5mo ago
It highlights the extremes the anthropic team adopts LLMs in their workflow. I think most of us live somewhere in the middle, using the right tool / output for the job.
15.
by adam_patarino 5mo ago
Sometimes I think Valve is just showing off
16.
by adam_patarino 5mo ago
Diamond hands?
17.
by adam_patarino 5mo ago
Most workforce reductions are using AI as a cover up for greedy short term bonuses. Any exec using AI to pay fewer people lacks imagination.
18.
by adam_patarino 6mo ago
The explanations are all fine. But they come after the team gaslit everyone, telling us it was a skill issue.
19.
by adam_patarino 6mo ago
It’s certainly #2. They have shown over dozens of decisions they move very quickly, break stuff, then have to both figure out what broke and how to explain it.
20.
by adam_patarino 6mo ago
Token anxiety is real mental overhead.
21.
by adam_patarino 6mo ago
Prices are not just hard cost of inference. Training costs are not equal. Chinese labs have cheaper access to large data centers. I also suspect they operate far more efficiently than orgs like openAI.
22.
by adam_patarino 6mo ago
Either they vibe coded a test that was extremely broken. Or they vibe wrote some bullshit to try and back pedal.
23.
by adam_patarino 6mo ago
What are you guys subscribed to if not Claude? Copilot? Or is everyone legit bringing their own license?
24.
Compressing LLMs with progressive pruning and multi-objective distillation(rig.ai)
4 points by adam_patarino 6mo ago|2 comments
25.
by adam_patarino 6mo ago
Compressing a mixture of experts model to fit on smaller hardware with a reinforcement learning approach called Self-Distillation Policy Optimization, progressive expert pruning, multi-objective knowledge distillation, speculative decoding,
26.
Teaching a Model to Code(rig.ai)
6 points by adam_patarino 6mo ago|1 comments
27.
by adam_patarino 6mo ago
Discusses the difference between training for one-shot / one turn coding tasks and agentic long horizon agentic coding workflows.
28.
by adam_patarino 6mo ago
Fluorographane is that stuff in factorio space age, no?
29.
by adam_patarino 6mo ago
Anthropic has had the chance to explain what they did rationally. Instead they chose to be opaque and grandiose. Giving them the benefit of the doubt is no longer appropriate.
30.
by adam_patarino 6mo ago
I did not enjoy this

More ›