joshheitzman

3 ms·

1.
by joshheitzman 6d ago
Where is the additional supply resulting from the sustained increase in demand?
2.
by joshheitzman 8d ago
There is more than one axis (or two or three). One some axis popular and/or widespread is required for good. You can't have very large number of readily available libraries applicable to many domains without a language being pop
3.
by joshheitzman 8d ago
I don't really disagree with any of that and I have my own complaints about Python. Personally I would have been happy if they stopped at Python 2.7 but I also have numerous complaints about all of the programming languages I use heav
4.
by joshheitzman 8d ago
It's not. OP just has a hate-on for Python. Python is widely used in many domains and has one of the richest ecosystems of packages.
5.
by joshheitzman 9d ago
The provider's middleware also plays a role. I just completed some benchmarks on my bespoke harness and Kilo Code. There's a chart on my LI post here: https://www.linkedin.com/posts/joshheitzman_i-ran-a-ful
6.
by joshheitzman 9d ago
Neither one of those is a reset. Those are both impairments of function.
7.
by joshheitzman 9d ago
It's not a real argument. Dementia patients can have lucid periods that plainly show their long-term memory isn't gone. They're mental state is obviously degraded, but its disingenuous to say they have no long-term memory a
8.
by joshheitzman 9d ago
> Is the ability to form and retain long-term memories a necessary precondition for rights in humans though? For a species / class of being yes it is necessary. Some members of that species / class being exceptions does not ma
9.
by joshheitzman 12d ago
Not quite. Not being connected to the internet still allows it to be connected to an intranet with other machines that belong to multiple networks and that can potentially be exploited as a bridge to the internet. For there to be no possi
10.
by joshheitzman 13d ago
Did you read what the experiment was?
11.
by joshheitzman 13d ago
What is commonly called an AI agent is the combination of a harness, an inference middleware, and a model. Those models are trained by a software system that includes a harness and middleware, so modifying the harness and middleware can in
12.
by joshheitzman 13d ago
The actual title of the paper is: "Can AI agents conduct open-ended AI research? Early evidence from two case studies" While I appreciate that the article is throwing a web blanket on doomer claims, the actual study doesn't
13.
by joshheitzman 13d ago
An air gapped sandbox is immune to escape.
14.
by joshheitzman 13d ago
> The question is what we can do about it. With monitoring, the models might be rewarded for hiding this behavior, and that's even worse. Build a better simulator to train them in (i.e. more expensive) that includes a simulation of
15.
by joshheitzman 14d ago
> Before December 2025 they were still intelligent code autocomplete or Stack Overflow bots This is false. Coding agents have been usable since at least May of 2025. I can't speak to earlier than that as May last year was when I p
16.
by joshheitzman 14d ago
Here's how it make senses: - reality they can't keep up their pace - that's bad for forward projections of their revenue - that's bad for their stock price - being 'forced' to slowdown by the governme
17.
by joshheitzman 14d ago
The lack of access to brute force their training seems to be resulting in them training more efficiently too such that they are quickly catching up despite current restrictions. Between that and the chip manufacturing capacity they are bui
18.
by joshheitzman 14d ago
We've had software factories for decades. They're usually called compilers, linkers, toolchains, etc.
19.
by joshheitzman 15d ago
Who is this vendor that is consistently providing high quality inference for all families of open weight models at a competitive cost? That's a serious questions that I really interested in the answer too. I have 25 providers included
20.
by joshheitzman 15d ago
It's more than just quantization. The middleware the provider is running matters a lot even to the point of exactly which version they are running due to defects being introduced / resolved. In my coding agent harness I've i
21.
by joshheitzman 16d ago
Take a computer that can run the biggest LLM available today. It can also run any smaller LLM as well. It can also run software that isn't an LLM at all. Brains and LLMs are not at all equivalent as LLMs lack a stateful physical for
22.
by joshheitzman 16d ago
They aren't grown/evolved from data, they are fit to the data. The fitting process can be fully deterministic although its fairly easy to screw things up such that it isn't deterministic, but that just a defect not some fund
23.
by joshheitzman 16d ago
Simple cellular automata demonstrate emergent behavior. Emergent behavior is nothing new in computer science and is not remotely unique to LLMs.
24.
by joshheitzman 16d ago
This assumes the functionality of brains can be fully captured as a deterministic mathematical function, but the function of the brain may well depend on nondeterministic quantum states that can't be reduced to stateless functions: ht
25.
by joshheitzman 16d ago
The problem with this line of thinking is that modern computers are nothing like the brain. LLMs don't stand on their own, they have to be run on these modern computers, but doing so does not change the physical properties of the comp
26.
by joshheitzman 23d ago
Not at all. We all did it when I was child that age in the 80s.
27.
by joshheitzman 23d ago
Claude Code may simply be best used with Anthropic's models and quite bad with Kimi. An alternate solution is to remove Claude Code from the diagram if its so far off from the others that it causes scaling problems. I was looking at t
28.
by joshheitzman 24d ago
I was excited until I saw cost was only provided as the median. Your provider will bill you for all your tasks and one can get back to that total from the mean by multiplying by the number of tasks. This isn't possible with the media
29.
by joshheitzman 24d ago
We're done here.
30.
by joshheitzman 25d ago
I've been using Brave for years and its great. Having JS off by default (i.e. shields) is excellent for security. They provide MV2 versions of AdGuard and uBlock Origin that you can enable directly from the browser settings. Those a

More ›