∀𝗥𝕀千

@LottoLabs @itgoatee If it is centrally placed, you can use it as a heat pump during winter 😁

Woah 7x3090s for sale $6600 obo @itgoatee put me on to this ad reddit.com/r/homelabsales…

@TheAhmadOsman There is bit of a knowledge gap on how to run models locally. The number of params (llama.cpp) can get quite complicated.

You should buy an RTX 3090 and learn how to run models locally The elite don’t want you to know this but running local models is hella easy, performant, and cheap nowadays

@ProudSocialist That’s true. AI doesn’t get frustrated. You do.

Tech Billionaire Marc Andreessen to Joe Rogan on why AI bots are better than human workers: "The bots never get frustrated with you…AI also never gets drunk. Never gets sick. Never gets depressed because his girlfriend broke up with him. Never files HR complaints." You don’t hate these tech billionaires enough. Humanity better rise up now while we still can.

@hunvreus Nah. Vibe coders are the real engineers.

@ayayalar I know, right? Should be using waterfall, UML and SAFe, otherwise they're not a real Engineer™.

Why on earth would you want to revert to Spec-Driven Development? Yes, agents are way faster at writing code. And (some) humans are better at system thinking. But we also suck at planning. Any experienced engineer knows you simply cannot sit down, write the specs and then write the software that matches it. At least not if you plan on writing something "good". You need to work through the problem to understand its boundaries and shape a solution that makes sense. Just leverage the fact that writing code is cheap: 1. Prototype, 2. Document learnings, 3. Rewrite based on learnings, 4. Document solution, 5. Refactor, 6. Document changes. Even if you have to repeat parts or all of this, you'll get to a good solution faster than with SDD.

@AlchemyAmerican I don’t think he did. He’s been saying this all along. He wants to see the proof.

The guy did a 180 degree flip on aliens because the department of War created a website and a documentary with high ranking intelligence officials came out. He was never doing real science

@Sevdamzz_GS Her şey tadında güzel

🚨İcardi konusunda tarafını seç. ▪️İcardi kalsın ▪️İcardi gitsin

@Fikirtepeli_83 Hala ağlıyo

Tanju Çolak Galatasaray dan Fenerbahçe ye gidişini anlatıyor "Ağlaya ağlaya Fenerbahçe ye gittim"

@UnslothAI I can’t fit Qwen3.6-27B-UD-Q5_K_XL.gguf with 128k ctx in rtx 5090. Am I doing something wrong or that’s expected?

4-bit Qwen3.6 MTP GGUF managed to search 70+ sites from a single prompt. Try this locally on 20GB RAM via Unsloth Studio. Unsloth now supports auto MTP + speculative decoding & auto-selects the best MTP settings for your device (Mac, CPU, GPU). GitHub: github.com/unslothai/unsl…

@neetcode1 I give this guy no more than six months before Microsoft let him go.

So he’s saying Microsoft will be dead? Idk why no one ever talks about the second order effects of anything

@MemoryReboot_ If I am spending 10k, im buying rtx pro 6000

What's the point of buying a DGX Spark for $5k? Imo to fully unlock the potential you need at least two, that's already $10k Mac Studio M3 Ultra 256GB for $7.5k looks like the better play. It's faster and all in one box The only reason is that it just looks cool af What am I missing?

@cline wayyy too many issues. Unusable at this time. Will go back to the older version.

@cline I can confirm this is fixed!😀. There are lots of other issues though. I will keep updating to the latest as it becomes available.

@ayayalar we should have fixed it in the latest release v3.0.3. Please give it a try!

@cline tui is broken. Cannot register local openai compatible provider.

@Truthpole Corbell gets all the videos recorded by a high resolution potato 🥔

🚨 New - Some clips of Newly released UFO Footages from Jeremy Corbell's New UFO Documentary "Sleeping Dog" Credits - Jeremy Corbell

@RealAlexJones 2nd drop of files coming soon. May 4th 2056

🚨HAPPENING NOW: As Predicted, White House Releases Absolute NOTHING BURGER With Supposed Full Alien/UFO Disclosure! Alex Jones Breaks Down Exactly What's Really Happening & The Bottom Of The Alien Rabbit Hole! 🔴WATCH/SHARE ALEX JONES LIVE NOW: x.com/i/broadcasts/1…

@addyosmani We can deny all day long but the software engineering is going to be dead in the near future. It is a revolution and we’ll see what comes next.

Democratizing a skill never kills the craft. It just raises the floor and the ceiling. Software still has a promising future.

@mattpocockuk I love plans. Works great. I’d suggest stop influencing and let everyone decide for themselves.

The more I replace plans with prototypes, the better the outputs Who'd have thought that low fidelity prototypes were better than walls of spec Oh yeah, the entire industry for 20 years Stop going against decades of knowledge because someone in SF shipped it as a 'mode'

@zoranh75 Software design is important when people are writing the software. If AI is the writer and the maintainer, does it matter single new state requires changing dozen methods? Btw, I agree with you but I started questioning the idea of the importance of patterns like DDD.

You ask an AI tool to model a business entity, and in seconds you have a class with a handful of properties, an enum for status, and nulls scattered all around. The code compiles. Tests pass. And six months later, adding a single new state requires changing a dozen methods and, if you are lucky, only three validators. The tool generated what looks to be a working code, but what it gave you was not software design. These are not random mistakes. They are the predictable result of tools such as Claude Code and Codex, that generate code without modeling the boundaries between primitives, domain concepts, and states. In my latest video, I walk through AI-generated C# code and show exactly where these failures appear and why they matter. - Primitive obsession – Using `decimal` for an amount and `string` for a currency code feels natural until you realise that every caller must repeat the same validation. - Data clumps – When two or more values always appear together, their pairing is a concept the code has failed to name. - Enum-for-state – An enum carries no associated data, so every state that needs extra information forces nullable properties onto the containing class. - Impossible states made possible – When an object exposes a public method regardless of whether it is defined for the current state of the object, you know the type system is lying to callers. - The version 2.0 problem – A codebase built on these patterns can reach version 1.0 easily. Getting to version 2.0 is where the damage shows. These failures are not hypothetical. They appear in every codebase that accepts AI output without applying domain knowledge and design judgment. Watch the full video "Claude Code is Resurrecting Our Worst Nightmares" youtu.be/cPoEFM1AKE4 And, if you wish to learn more about object-oriented design and domain modeling, these are the learning resources I can recommend first-hand: 𝘿𝙤𝙢𝙖𝙞𝙣-𝘿𝙧𝙞𝙫𝙚𝙣 𝘿𝙚𝙨𝙞𝙜𝙣 by Eric Evans ► #ad" target="_blank" rel="nofollow noopener">codinghelmet.com/go/book-domain… 𝘿𝙤𝙢𝙖𝙞𝙣 𝙈𝙤𝙙𝙚𝙡𝙞𝙣𝙜 𝙈𝙖𝙙𝙚 𝙁𝙪𝙣𝙘𝙩𝙞𝙤𝙣𝙖𝙡 by Scott Wlaschin ► #ad" target="_blank" rel="nofollow noopener">codinghelmet.com/go/wlaschin-fu… 𝙁𝙪𝙣𝙘𝙩𝙞𝙤𝙣𝙖𝙡 𝙋𝙧𝙤𝙜𝙧𝙖𝙢𝙢𝙞𝙣𝙜 𝙞𝙣 𝘾# by Enrico Buonanno ► #ad" target="_blank" rel="nofollow noopener">codinghelmet.com/go/buonanno-fu… 𝙈𝙤𝙙𝙚𝙧𝙣 𝙎𝙤𝙛𝙩𝙬𝙖𝙧𝙚 𝙀𝙣𝙜𝙞𝙣𝙚𝙚𝙧𝙞𝙣𝙜 by David Farley ► #ad" target="_blank" rel="nofollow noopener">codinghelmet.com/go/modern-soft… 𝘽𝙚𝙜𝙞𝙣𝙣𝙞𝙣𝙜 𝙊𝙗𝙟𝙚𝙘𝙩-𝙊𝙧𝙞𝙚𝙣𝙩𝙚𝙙 𝙋𝙧𝙤𝙜𝙧𝙖𝙢𝙢𝙞𝙣𝙜 𝙬𝙞𝙩𝙝 𝘾# at Udemy ► codinghelmet.com/go/beginning-o… 𝗗𝗶𝘀𝗰𝗹𝗼𝘀𝘂𝗿𝗲: This post contains affiliate links. If you make a purchase through those links, I will earn a small commission at no extra cost to you.

@sudoingX 40 tp/s on qwen 3.6 27b was your dream. Tell us about the reality.

a week with the dgx spark, here is what is on it and what i have measured so far. nobody is really talking about this machine and it is quietly becoming the workhorse of my whole stack. hardware: nvidia gb10 sm_121, 124 gb unified lpddr5x at 273 gb/s, cuda 13.0 models on disk (305 gb total, 9 ggufs): > qwen 3.6 27b q4_k_m / q5_k_m / q8_0 / ud-q4_k_xl > nemotron 3 omni 30b-a3b q4_k_m / q8_0 / ud-q6_k / ud-q6_k_xl > deepseek v4-flash 158b q4_k_m (112 gb, flagship 128gb-tier test) terminal + shell environment: > zsh + oh-my-zsh + powerlevel10k theme > modern cli stack: bat, eza, ripgrep, fd, git-delta, tldr, neovim, fzf, autojump > 6 tmux sessions actively running for parallel agent work ml + agent stack: > llama.cpp built sm_121 against cuda 13 > uv + venv ml stack with pytorch 2.11.0+cu130 (aarch64) + transformers + diffusers + accelerate > hermes agent v0.11 with codex auth bridge > opencode for free-model overnight research > telegram gateway routing to nemotron q8 right now speeds verified so far: - nemotron 30b-a3b q8: 56 tok/s gen, 1,300 tok/s prefill, 96% gpu, 33gb in unified - qwen 27b dense q4: 40 tok/s consistent 90+ gb of unified memory still free. deepseek v4-flash 158b loading next as the real flagship test, multimodal omni testing once mmproj pulls, comfyui install in flight for the diffusion lane. honestly curious what the actual limit is on this box, i have not hit it yet.