seamossfet

5 ms·

1.
by seamossfet 8d ago
If you want to do a 1-bit model you have to QAT at pre-training with way more data than chinchilla to compensate for the cliffs (like 50x). Quantization on an existing pre-trained model will almost always collapse at 1-bit
2.
by seamossfet 14d ago
We use the ONNX runtime for small models in the browser https://github.com/microsoft/onnxruntime
3.
by seamossfet 20d ago
just switched to doom emacs this month, absolutely loving it tbh
4.
by seamossfet 21d ago
yeah, but I mean even prose specific claude-isms aside; the information itself is a weird patchwork of concepts
5.
by seamossfet 21d ago
I notice a lot of these AI written articles share this pattern where they'll present idea 1, then idea 2, and finally idea 3 which is some amalgamation of idea 1 and 2. Claude especially will present hybrid options and compromises to a
6.
by seamossfet 25d ago
Has anyone read this yet? I'm extremely skeptical of any book published after ~2024 just because of how much AI slop is in physical books these days. Looks interesting, but I'd like to see some reviews on the content first before
7.
by seamossfet 26d ago
I think a part of this is that people tend to undervalue their own skills and expertise when talking about these anecdotes. A lot of people in the comments do have a software engineering background. People at different skill levels in diffe
8.
by seamossfet 2mo ago
I don't know if I agree with the premise that having access to AI results in dulling human intellect. I feel like to get to Terry's level you need a combination of passion and aptitude for the subject. People that don't want
9.
by seamossfet 2mo ago
The big take away for is the fact that the ONLY reason why chatgpt was able to get to this counterexample was because of the knowledge of the person driving the conversation. I don't think chatgpt could have come to this on its own wit
10.
by seamossfet 2mo ago
Nothing wrong with vibecoding a toy project.
11.
by seamossfet 4mo ago
We're building a CAD for drug design, we often have to handle large and highly varied file formats. Protein structures, compounds, python scripts, lab notebook entries, instrumentation data, etc. From a data structure and file ergonomi
12.
by seamossfet 4mo ago
Does this provide gitflow to handle conflicts from multiple agents touching the same file system or is it purely for single-branch sequential iterations on the filesystem? I have a use case that could use this if it supports handling branch
13.
by seamossfet 5mo ago
People often use "non-deterministic" to imply "unreliable" when that's not always the case. You can have an LLM that's clearly non-deterministic but also reliable for tasks.
14.
by seamossfet 5mo ago
Great article, but your site background had me trying to clean my laptop screen thinking I splashed coffee on it.
15.
by seamossfet 5mo ago
The hardest part about using agents to code for me has always been working in teams. When you can cut through huge parts of the code with a chainsaw, how do you review multi-thousand line PRs? It's really hard to do surgical changes wi
16.
by seamossfet 5mo ago
This is why I built [AI slop tool]. [Self promotion link to my vibe coded startup with no users]
17.
by seamossfet 5mo ago
A false dichotomy that segments typical replies into one of two groups. Group 1: A thinly veiled straw man that buckets everyone I disagree with, along with an attempt to appear as if I'm being unbiased Group 2: The group I put myself
18.
by seamossfet 6mo ago
The problem with models like this is they're built on very little actual training data we can trace back to verifiable protein data. The protein data back, and other sources of training data for stuff like this, has a lot of broken str
19.
by seamossfet 6mo ago
This is awesome! The only limiter here is the resolution, I think this is fantastic for cellular level organelles but it doesn't quite get down to the same resolution something like x-ray diffraction does. There's a huge trade off
20.
by seamossfet 6mo ago
Honestly, this is a good thing. OpenClaw as a concept was rather silly to run such a heavy model for. If you want something like OpenClaw to work you really need to figure out how to do it with an economical model.
21.
by seamossfet 6mo ago
Honestly, it'd be really funny to try and make a CLAUDE.md file for slop maxxing.
22.
by seamossfet 6mo ago
I'm not convinced people who are doing real work on production applications with any sizable user base is writing code through only agents. There's no way to get acceptable code from these models without really knowing your code b
23.
by seamossfet 6mo ago
Oh my god, this comment gave me flashbacks to when I was writing android apps in Eclipse + ADT
24.
by seamossfet 6mo ago
> We very much still believe this That's good to hear, I might have jumped a little too quickly in my opinion. It's a bit of a Pavlovian response at this point seeing a product I very much love embrace a giant chat window as a
25.
by seamossfet 6mo ago
Yeah that's the disconnect though right? Even with the best frontier models, you need to do a lot of system design work, planning, and reviewing before you can let these models run. These models are infinitely more effective when pilot
26.
by seamossfet 6mo ago
Man, I wish they'd keep the old philosophy of letting the developer drive and the agent assist. I feel like this design direction is leaning more towards a chat interface as a first class citizen and the code itself as a secondary conc
27.
by seamossfet 6mo ago
I wonder how much of this is also used for audience segmentation for their advertisements? Linkedin ads are some of the most expensive out of any social media platform, but they also tend to have the highest conversion since you can get pre
28.
by seamossfet 6mo ago
>But if your workload is not shifting from write-heavy to read-heavy, you inevitably will be responsible for a major outage or quality issue. I think that's actually a good way to look at it. I use AI to help produce code in my day
29.
by seamossfet 6mo ago
>slop cannons I am stealing that phrase haha
30.
by seamossfet 6mo ago
Yeah, I'm not trying to defend slop. I don't think all means-to-end people are just in it for money, I'll use the anecdote of myself. My team is working on a CAD for drug discovery and the goal isn't to just siphon money

More ›