smcleod

7 ms·

151.
by smcleod 8mo ago
Yeah I think it's either a billing bug, or some sort of inbuilt background sub-agent loop gone wild inside Claude Code, if you have a look at recent issues on the Github relating to 'limits', 'usage', 'tokens&#
152.
by smcleod 8mo ago
It hasn't always done this, it's a relatively recent problem in the last 1-4 weeks (roughly). (note: I'm on the $160AUD/mo plan, so I think that's $100USD).
153.
by smcleod 8mo ago
Considering all the problems they've been having with over-charging Claude Code users over the past few weeks it's the very least they could do. Max subscribers are hitting their 5 hour usage limits in 30-40 minutes with a single
154.
by smcleod 8mo ago
I don't think much from OpenAI can be trusted tbh.
155.
by smcleod 8mo ago
Microsoft's everything is running into problems. Seriously though, it always seems like a dumpster fire over there but the last year is next level with all their Windows 11 problems, Windows update problems, Office365 outages, Copilot
156.
by smcleod 8mo ago
Sounds like they've been using skills incorrectly if they're finding their agents don't invoke the skills. I have Claude Code agents calling my skills frequently, almost every session. You need to make sure your skill descrip
157.
by smcleod 9mo ago
Re: Flynn effect reversal - The article attributes this entirely to classroom technology. Researchers have proposed multiple explanations including changes in education, nutrition, cultural factors, and media exposure. The researchers study
158.
by smcleod 9mo ago
Came to say this as well, usually they reduce manufacturing of the old model as the new one ramps up.
159.
by smcleod 9mo ago
There's also smaller models / lower context variants for things like title generation, suggestions etc...
160.
by smcleod 9mo ago
https://smcleod.net
161.
by smcleod 9mo ago
Tax large companies properly then you don't have to tax the public for things like this.
162.
by smcleod 9mo ago
I always enjoy watching your videos Jeff, appreciate your honest, no-hype approach. As a bonus it's great to see you active here and on Reddit as well.
163.
by smcleod 9mo ago
MS Windows only unfortunately!
164.
by smcleod 10mo ago
Need the M5 (max/ultra next year) with it's MATMUL instruction set that massively speeds up the prompt processing.
165.
by smcleod 10mo ago
This was an interesting read, thanks for sharing. I've recently been building something that uses Parakeet v2/v3 models, I'm using the parakeet-rs package ( https://github.com/altunenes/parakeet-rs ) which
166.
by smcleod 10mo ago
My 2023 Macbook Pro (M2 Max) is coming up to 3 years old and I can run models locally that are arguably "better" than what was considered SOTA about 1.5 years ago. This is of course not an exact comparison but it's close enou
167.
by smcleod 10mo ago
I suspect Ollama is at least partly moving away open source as they look to raise capitol, when they released their replacement desktop app they did so as closed source. You're absolutely right that people should be using llama.cpp - n
168.
by smcleod 10mo ago
On a $20/mo plan doing any sort of agentic coding you'll hit the 5hr window limits in less than 20 minutes.
169.
by smcleod 10mo ago
The m5 max should be out around feb - wait till then
170.
by smcleod 10mo ago
Indeed! I checked their status page within 2 minutes of having issues and it was updated to show they had detected it.
171.
by smcleod 10mo ago
No I don't believe so. Cursor is usually pretty behind other agentic coding tools in my experience.
172.
by smcleod 10mo ago
They probably used Amazon Q to generate them...
173.
by smcleod 10mo ago
The main jump is from anything to M5; not because it's simply the latest but because it has matmul instructions similar to a CUDA GPU which fixes the slow prompt processing on all previous generation Apple Silicon chips.
174.
by smcleod 10mo ago
The DGX Spark is not good for inference though it's very bandwidth limited - around the same as a lower end MacBook Pro. You're much better off with a Apple silicon for performance and memory size at the moment but I'd recomm
175.
by smcleod 10mo ago
Oh that makes a lot of sense, I'm glad it works that way actually - the explanations afterwards left me wondering if it was truly explaining the connections or if it was inferring what they would be (leading to a problem a bit like how
176.
by smcleod 10mo ago
That's quite fun, I wish it had more information about which model took that path and the inference / sampling parameters.
177.
by smcleod 10mo ago
I see it requires an ANTHROPIC_API_KEY, does that mean it does not work with Claude Max plans?
178.
by smcleod 10mo ago
Always enjoy reading your blog Simon!
179.
by smcleod 10mo ago
This one is a bit old now so a number of things have changed (I mostly use Claude Code now, Dynamic context (Skills) etc...) but here's a brief TLDR I did early this year https://www.youtube.com/watch?v=dDSLw-6vR4o
180.
by smcleod 10mo ago
1. I didn't say it was a best example, I replied to a comment asking me to "Post a repo" - I posted a repo. 2. Straw man argument. I was asked for a repo, I posted a repo and clearly you didn't look at the code as it&#x2

More ›