
Juan Cruz Martinez
6.7K posts

Juan Cruz Martinez
@jcmartinezdev
Deep dives on engineering careers and AI’s real impact on developers. Writing The Long Commit on Substack. 20 yrs building software. @Auth0


Devs hating on AI and saying it's going to fall and it's a bubble that's not useful for coding reminds me of: - People hating on cars saying horses are the best. - People hating on electricity? - People hating on <put innovation here>



Thank goodness our trade deal with the US was rendered null and void because we would’ve had to deal with the availability of free-roaming grass-fed Brazilian beef supply in Malaysia being slowly replaced by beef from disgusting American factory-farmed feedlots.

New: Readwise for @openclaw Save anything with one click from your browser or phone: articles, tweets, books, youtube, podcasts, newsletters, more. Your claw now has instant access to that full library as markdown. Easily search and actually stay on top of it all.


You have until 24 April to do it, and then *boom* you're now part of training their models. GitHub is truly losing its identity and what it stands for. Opting in paying customers means they no longer see themselves as the infra for code. Opportunity for someone else!

🎙️Do you know you now have all the building blocks for full speech-to-speech? - Voxtral Realtime: High-quality, real-time speech-to-text. - Mistral Small 4: Fast, efficient, general-purpose agentic model. - Voxtral TTS: Realistic customizable text-to-speech with streaming output.



building reliable agents comes down to just this one tip: scripts own facts. agents own thinking deterministic control flow + cognitive agents = reliable learned this from zoe reviewing her own pipeline, where she wrote: "Zoe should not be trusted with control-plane facts." self-aware queen





mac dilemma update (73 comments later): going with M4 Max 128GB. now. here's why: 1. opportunity cost - every day my agents are bottlenecked is a day I'm not shipping (wwdc in june is 5 AI years away) 2. pre-AI pricing won't last - apple priced these before AI agent demand 3. vertical scaling = focus - managing network I/O across multiple minis is time I'd rather spend building 4. local = control - no rate limits, no API outages, full autonomy 5. memory > compute - RAM is the bottleneck for agent swarms, not CPU my plan now is to use the studio to power my agent swarm but rely on cloud models (codex 5.3 mostly), but one comment that stuck with me from @HEXtheantidote: "running a local model is the only way for them to achieve true autonomy" local models aren't there yet. but in 12 months? having a self-hosted AI might be as normal as having a router. Mac is the body. agent is the soul.
















