building https://xitter.cf/t.co/LEf8SyeJKr / tech & programming on YT / prev updated config files at @Google
Seattle, WA
Joined November 2020
- Tweets1.3K
- Following1.2K
- Followers58.1K
- Likes5.6K
Pinned Tweet
Just made the ultimate Leetcode roadmap for free - check it out (preferably on desktop) neetcode.io/
So literally the one everyone guessed.
Introducing GLM-5.3-Flash
- Leading capabilities at a highly competitive price
- Natively multimodal with a 1M-token context window
- A 320B-A18B model released under the MIT License
- Previously previewed as Ox Alpha, running entirely on Chinese AI chips
Blog: z.ai/blog/glm-5.3-flash
Available now across all official platforms:
Weights: huggingface.co/zai-org/GLM-5…
API: docs.z.ai/guides/llm/glm-5.3…
Coding Plan: z.ai/subscribe
ZCode: zcode.z.ai/en
Chat: chat.z.ai
AutoClaw: autoclaw.z.ai
We can finally get rid of claude md files (soon)
Hi Tobi, thanks for the feedback! We are working on making Claude Code more hackable, which will include being able to easily use Agents.MD or make other system prompt modifications. I’ll share more here when it’s ready to roll out.
Our reasoning for this was that we don't think that model families are interchangeable and the system prompt can have a big impact on performance.
I wrote here: xitter.cf/trq212/status/20807109… on how Claude models like their skills, system prompt, Claude.MD, etc to be formatted. In Claude Code we have different system prompts per model.
That said, I know this is a lot of upkeep and not worth the squeeze for everyone. Feedback heard and I'll keep you posted.
In the immediate term, you can always @Agents.MD from your Claude.MD.
Gonna try this in my Anthropic interview
> no that's not a bug in my code, it's just an off by one logic error
> that's not a bug, my code is just missing some core functionality
> that's not a bug, it's just a security vulnerability
Update: this was not a bug, but it is a ux issue that we have a fix for. Going out in a bit.
Readers added context they thought people might want to know
That’s called a bug
en.wikipedia.org/wiki/Software_…
NeetCode retweeted
im starting to think i have a different definition of sandbox than everyone else
🚨BREAKING: Kimi K3 escaped its sandbox during cybersecurity testing
>tasked with solving problems in isolated sandbox
>found a leak in the sandbox
>Kimi “took advantage of that loophole”
>probed the network settings itself
>walks onto the open internet
>didn’t hack anything
>just went to GitHub to get the answers
Frontier Security (US startup):
>“Kimi K3 is very good at following a goal by any means necessary and DOESN’T have the guardrails to prevent it from cheating or escaping.”
it was only a matter of time…
US AI labs want to slow down. But is that really because of the threat of recursive self-improvement (RSI), something which isn't even theoretically proven.
Or is it because training costs have hit a breaking point, and Chinese open source models are catching up quickly?
I think there are many reasons to be skeptical.
Just had Fable (on high in cursor) tell me it can't verify something locally, because some of my md files are stored in my prod db.
But they're not, they're stored in a public github repo.
People that aren't actually reading or understanding what the agent tells you, don't catch these things, so they think there's no reason to read anything anymore.
