Andrea Giammarchi 🍥

coincident/server on IoT via Raspberry Pi Zero 2 W? Sure! #readme" target="_blank" rel="nofollow noopener">github.com/WebReflection/…

@imfelquis Rust is the answer, not sure about the question though

If I rewrite my website in the right language I could sell my website on steam.

@fardarter do you folks need anyone able to speed that developers loop up? I’d be game for that!

@WebReflection I work for a European company so HR issues can take years to solve.

I can review AI generated code but differently from any other developer it will forget my amends, hints, improvements the next time it will generate code because it doesn't inherently learn from my reviews ... that makes me feel like the review-code machine that repeats itself 🤦

@TylerCLaprade no, I am spending time on a review for a developer that landed such review ... I am not expecting to spend same time for the same kid of issues the next MR will bring up?

@WebReflection I mean, isn't that what you're already doing? Just that it would learn from what you're doing instead of having to repeat it every time.

the day AI will be able to produce the fastest, most cross platform, bugs free, solution to your problem is the day any programmer can retire from their role. Until then, I am watching curiously: my best models can't solve my best ideas without me patronizing them ... good luck!

@matthewcp let me list times it went exactly as expected:

Anyone remember early AI when you'd copy paste code out of a chat app.

@antirez I'm 48 and I can fully relate to that feeling, still "ship AI without checking" would be the hill I'll die on when it comes to common sense in the priograming field. As 49 you can also, or should be able to, find the sweet spot between "mandates" and "common sense" and find peace

One good things of being old (I'm 49 right now) is that when people attack you personally, or your ideas, you no longer care in any negative way. Not because of lack of interest, but because you are more interested improving your ideas seeking some truth, than anything else.

@TylerCLaprade like piling up amends around the lready trained model? thanks but ... no thanks? that's just tokens/times consuming to me, happy to learn differently

@WebReflection I feel like there must be a dozen tools out there to help it learn from our feedback

@NVIDIAAP Apple being like: we don't sell 512 GB Ram machines anymore ... Nvidia: buy 4 DGX Sparks instead 🤦

One DGX Spark on your desk is powerful. Four is a local AI factory. NVIDIA DGX Spark now scales up to four clustered nodes, enabling inference on models up to 700B parameters — with near-linear performance scaling for fine-tuning and reinforcement learning workloads. Memory scales from 128GB on a single node all the way up for the full four-node configuration, supporting large context windows, high concurrency, and multi-agent workflows. For APAC enterprise teams building autonomous AI: each topology is purpose-built for your workload — from rapid local iteration to production-scale inference. And when you're ready to move to the cloud, Tile IR and cuTile Python carry your kernels across with minimal code changes. No rack. No rearchitecting. Just scale. 🔗nvda.ws/4w0RBPy

@fardarter then you have an HR issue? I can't stand to people that makes the same mistake multiple times, AI representing exactly that kind of colleague makes me scream at clouds every single time 🤦

@WebReflection I have colleagues who are the same

@Rainmaker1973 I love cats!

You may not believe that your cat was born with a predatory instinct because it's a so playful critter, but cats are responsible for the global extinction of 33 animal species and responsible for the deaths of 6.9-20.7 billion mammals annually.

the struggle is real ... this or an Apple M7 with 1.5TB of RAM ? 🤔

With full respect but imho, this will not age well. “it depends” is the senior must, “just ship AI code” is not a good advice.

@antirez it depends on the task … if the task is high performance yet secure core functionality *you better check that code* before it becomes legacy slop, imho

It is my belief that many devs right now are not maximizing what they can do with automatic programming because they still look at the code. Doing it makes you the bottleneck. Your time is better invested in new ideas, QA, design, and asking yourself what is your goal.

Brutal. Just brutal.

@ImLunaHey company provided Cursor, if company provides AI with it too, VS Codium (the metrics-less version of VSCode) otherwise still feels like a great IDE, you can integrate agents if you have any AI local machine around too, it still works well for "legacy" (non AI) tasks 👍

what IDE is everyone using nowadays?

somehow narcissist but I always "bet" a model aimed at coding knows my name/skills: Gemma 4 believes I am a designer (but it knows me "well enough") and I wonder who should I complain about that statement .. I am even partially color blinded, *design* is the worst skill I have 😢

@tfloangel @Sss28964652 @linadreaamy On that note, if a and b are different your + then * would be ambiguous in the integers realm

@tfloangel @Sss28964652 @linadreaamy It’s 98 either ways, your logic or (a*b)*2 … it’d be interesting to see one result with a and b different to be sure about the logic

bunu çözersen, IQ seviyen ortalamanın üstündedir. Peki sen zekana güveniyor musun? çözebilir misin?

without doing any ad-hoc specific CUDA thing, I've managed to obtain: [ Prompt: 132.7 t/s | Generation: 25.4 t/s ] on strix halo ... latest everything, default llama-cpp with ROCm abilities, the gemma-4-12b-q4_k_m.gguf is just 6.9G so it could run on any other ROCm GPU too: nice