ojosilva

8 ms·

31.
by ojosilva 11mo ago
Here's a customer of the $200 max plan for 2 months. I fell in love with the Qwen3 Coder 480B model, Q3C, that was fast, twice the speed of GLM. GLM 4.6 is just meh, I mean, way faster than competitors, and practically at Sonnet 4.x
32.
by ojosilva 11mo ago
For the fans of the genre, some more speculation by the master speculator himself: https://avi-loeb.medium.com/post-perihelion-data-on-3i-atlas...
33.
Post Perihelion Data on 3I/Atlas(avi-loeb.medium.com)
1 points by ojosilva 11mo ago|0 comments
34.
by ojosilva 11mo ago
I've used it with CC and the match was great, not a lot of issues, I believe Qwen had a clear focus on distilling Anthropic models. GLM 4.6 is slightly better maybe, but the speed dropped to half on Cerebras so that's the price fo
35.
by ojosilva 11mo ago
Yeah, I'm just going through the Cerebras migration at the moment. It's a shame Cerebras completely dropped Qwen3 Coder's fast tool calling, short and instant responses, and better speed overall for GLM 4.6 thinking. Qwen3 is
36.
by ojosilva 11mo ago
I think the consensus in the future will be that LLMs were, after all, stochastic parrots. The difference with what we think today is that in the future we'll have a new definition of stochastic parrots, a recognition that stochastic p
37.
by ojosilva 11mo ago
I'm using CC + GLM 4.6 from Synthetic and results are top notch for $60/mo, speed is fast and servers are close to home than z.ai
38.
by ojosilva 11mo ago
The problem with Node is that the async decision is in the hand of the leaf node, which bubbles up to the parent where my code sits. Async/await is nice and a goal in most modern Node, but there are codebases (old and new) where async&
39.
by ojosilva 1y ago
I hope Claude Code follows the same path. It's just wrong to have such a key tool in your development toolchain obfuscated. Changes to CC reflect in the way I work. I've been modding the app for ever now. Something as simple as ev
40.
by ojosilva 1y ago
Yeah, I'm one of them, using Qwen 3 coder on Cerebras as coding agent through CC. What I keep hearing is (very ballpark anecdata)... 50% of people into coding agents are quite concerned about that last mile in difference with frontier
41.
by ojosilva 1y ago
I think MCP servers are valuable in several ways: - bundled instructions, covering complex iteractions ("use the id from the search here to retrieve a record") for non-standard tools - custom MCPs, the ones that are firewalled fro
42.
by ojosilva 1y ago
Their customer base are enterprise, so the issue can be addressed in private channels. There's little to be gained from making this particular breach public, from their point view. If anything, it's F5 customers who should advise
43.
by ojosilva 1y ago
I absolutely agree, Qwen Coder 480b is my go to, I'm also paying for OpenAI Plus/Codex and Cursor and none of those models ever give me an edge over Qwen. One can always setup an "ultrathink" model for when push-comes-to
44.
by ojosilva 1y ago
Yeah, that's really sad and totally undermines my UX on iOS (my iPad particularly). On my Android phone and macOS FF is my go-to browser, a delightful, irreplaceable experience. Sometimes people are amazed by the experience when I show
45.
by ojosilva 1y ago
And slapstick comedy Loaded Gun (1988) although that was a watch that would trigger the unsuspecting individual into a cold-blooded killer. I've warned about these poisoning scenarios not long ago and got called out for "fearmonge
46.
by ojosilva 1y ago
I think LLMs have what it takes at this point in time, but it's the coding agent (combined with the model) that make the magic happen. Coding agents can implement copy-pasting, it's a matter of building the right tool for it, then
47.
by ojosilva 1y ago
And I don't think you can revenue share these generations with rights owners just like that. What rights owner will let their "product" be depicted in any imaginable situation by any prompt by anyone in the planet? Words are
48.
by ojosilva 1y ago
Not sure but I think OP means the investors in OpenAI. Rephrasing: there are too many investors and orgs (ie MSFT) that want to "be in AI" and very few outlets that actually deliver the frontier hard core tech: Google, OpenAI, Ant
49.
by ojosilva 1y ago
“Do you think we’re going to die?” “No.” “We’re not?” “No.” “Why not?” “Because we’re carrying the fire.” “What fire?” “The fire inside. Goodness. That’s what we’re carrying.”
50.
by ojosilva 1y ago
Comments in code are instant technical debt. They need to be maintained alongside the code, so you are *programming" twice. Avoid comments, except when they really explain some obscure, incomprehensible section of code or to prevent ex
51.
by ojosilva 1y ago
To @simonw and all the coding agent and LLM benchmarkers out there: please, always publish the elapsed time for the task to complete successfully! I know this was just a "it works straight in claude.ai" post, but still, nowhere in
52.
by ojosilva 1y ago
In CS most security issues are treated separately from the fundamental engineering core. From the sw engineering standpoint the bridge is solid, if later some crooks can use it to extort users or terrorists can easily "make people fall
53.
by ojosilva 1y ago
You're right. That would be a "lethal double" then, a "lethal exacta" in horse racing. A trifecta is not needed for prompt injection to be dangerous.
54.
by ojosilva 1y ago
Sure! Yeah, definitely it's a coding agent zoo out there. But you can actually notice the polish in CC. Codex looks promising, a more naked look, I hope they invest in it, it's OSS and built in Rust, nice. Qwen has that same jumpy
55.
by ojosilva 1y ago
Opencode, been there, it has a Cerebras/qwen-coder out of the box preset! Unfortunately I found the TUI somewhat clunky, I dislike text windows and the terminal scrolling was not available, it was enclosed into the current viewport. On
56.
by ojosilva 1y ago
Claude Code has been an awesome experience for me too... without ever subscribing to an Anthropic account! I've never liked the free-tier Claude (Sonnet/Opus) chat sessions I've attempted with code snippets. Claude non-coding
57.
by ojosilva 1y ago
+1 for using the file system as memory. You don't need a memory backend for that, or MCP, RAG, nothing. Just the basic read/write fs tools any agent has at its disposal. Use an .agent/ directory. Tell it to keep a task list,
58.
Down and Out with Cerebras Code(infoworld.com)
2 points by ojosilva 1y ago|0 comments
59.
by ojosilva 1y ago
From your link: > In most cases, such attacks are discovered quickly and the malicious versions are removed from the registry within an hour. By delaying the infected package availability (by "aging" dependencies), we're
60.
by ojosilva 1y ago
I've been on both ends of this workflow. Sales always wins. "Wow. Well, I sure hope you don't get fired over this. Good luck. We'll scope it out and let you know how much time we'll need." "We'll see.

More ›