zach retweetledi
The OpenAI insider @thsottiaux has a warning for everyone offloading their thinking to agents.
English
zach
533 posts
zach
@battlerax
Father. Husband. Software Engineer building AI solutions.
Texas Katılım Mart 2026
250 Takip Edilen15 Takipçiler
It took me like 2 months, but I've grown to love gpt-5.5.
You have to prompt entirely different and put some time into your agents[.]md. Now that I'm over the hump, I can't really use any other model for code.
English
wish I had this on my last flight 😭
Elon Musk@elonmusk
Starlink coming to American Airlines! 💫
English
zach retweetledi
Running an NVIDIA DGX Spark connected to Starlink on a picnic table in the middle of the woods.
No office. No data center.
Just a supercomputer, a Starlink, and trees.
Vibe coding from anywhere on the planet is no longer a dream.
It's my setup right now.
English
zach retweetledi
zach retweetledi
(2x DGX Sparks) + MiniMax M2.7 NVFP4 = 16 local AI agents running simultaneously 👀
mr-r0b0t@mr_r0b0t
16 local AI agents streaming at once! MiniMax M2.7 NVFP4 — 2x GB10, no cloud APIs.
English
zach retweetledi
Totally worth it, you can run local models while doing a million other things. Apple 256GB when??
BridgeMind@bridgemindai
MacBook Pro M5 Max is fully set up and running local models. I have never seen speeds this fast. Qwen 3.6 35B and Gemma 4 31B are running blazing fast on 128GB of unified memory. Faster than both of my stacked NVIDIA DGX Sparks sitting right next to it. Initial impressions: Apple silicon in 2026 is no joke for local inference. The M5 Max handles 30B+ parameter models like they're nothing. Full review and comparison video coming soon.
English
@bridgemindai @matthewmillerai I’ve found that prefill is slower but tps is faster on MBP. Spark is kneecapped with low mem bandwidth, so dense models are painfully slow. MoE helps.
English
Just bought a MacBook Pro M5 Max with 128GB of unified memory for $5,399.
Picking it up today.
4TB of storage. Maxed out specs.
I want to see what local models can actually do on Apple silicon in 2026.
Testing everything against my two stacked NVIDIA DGX Sparks.
Same models. Same prompts. Real comparison.
Full results coming this week.
English
zach retweetledi
The problem with the statement “I don’t write code anymore” is everyone has a different version of what it means
For instance to me I don’t hand write code, but the LLMs are still essentially a really dumb scribe that if I didn’t read their code things would go south fast
English
Thinking about a series where I look at high quality skills and:
- break down why they work
- try them out
- tell you what you should steal from them
English
zach retweetledi
At this point I think OpenAI has more MacOS engineers than Apple does
OpenAI Developers@OpenAIDevs
Codex anywhere and everywhere, all the time. Now your Mac doesn’t have to be unlocked for Codex to use your computer. From your phone, Codex can securely use apps on your Mac, even when the screen is off and locked. #locked-use" target="_blank" rel="nofollow noopener">developers.openai.com/codex/app/comp…
English
Data poisoning starting to become present?
Tenobrus@tenobrus
i'm sorry WHAT DO YOU MEAN THE "HIDDEN TEXT"???
English
zach retweetledi
@robinebers It has been noticeably better for me as well! Pretty much immediately since announcing it
English
claude code limits have INCREASED and everyone screaming the opposite is wrong
how the fuck would I know?
well, I have a claude project running for about a year
stuffed with business context, almost at the limit
6 months ago my 5h window was constantly exhausted
now I barely use 13% in a full hour
so limits have definitely increased, yet everyone claims the opposite - why's that?
I don't really know, but I suspect that's because MCP servers, skills, goals, subagents, and agentic loops that burn more tokens than ever now
the limits got better
but the workflows got more expensive
English
zach retweetledi
