@melfie@lemmy.zip

comments

• “Stable genius” would be more apt because the exponential benchmark growth period appears to be over and the term has an ironic meaning from Trump’s own usage that very much applies to LLMs.

  on Trump says the US is officially renaming AI to ‘super intelligence’ in c/technology@lemmy.world · 6h ago

• I can attest this does work well. Having LLMs review your plans for how to do something and point out what you missed works well too. I also have my local LLM review creative writing without writing anything, give feedback about my Blender renders and other artistic things, and be a sounding board for ideas in general. Be warned that Qwen has no mercy and will shit all over your passion projects.

  on Engineer says Claude Code has made his job "soul-sucking" as workers spend 12-hour days pressing enter in c/technology@lemmy.world · 22h ago

• I have no idea if LLMs actually make me faster in the aggregate, but I’m more interested in using them to be better. For example, having a LLM review some code, or share a plan for and have it point out all of the flaws in your reasoning or alternatives you never considered.

  on Shopify CEO says employees' 'slop grenades' are making more work for everyone else in c/technology@lemmy.world · 2d ago

• In theory, an automated system should be able to outperform humans at a rote task like this under the right conditions. However, I don’t imagine throwing LLMs in the mix without redesigning the whole system for automation is going to end well.

  on FAA tees up $875M AI tool to help manage air traffic congestion in c/technology@lemmy.world · 3d ago

• With all the Russian piracy sites, I assume nobody in Russia actually pays for software. I figure they just haven’t gotten around to cracking that machine yet.

  on Putin casts online vote using unlicensed Windows on an American Dell PC in c/technology@lemmy.world · 4d ago

• I’m assuming a lot of its training data is synthetic and distilled from Chinese models that were themselves trained from pirated data and distilled from American models trained on pirated data. It would be quite remarkable if the training data involved no piracy whatsoever. Then again, it’s open source, so I suppose it would be essentially reversing a reverse Robin Hood.

  on is K2 horizons historically significant as the first actual open-source LLM, or am I falling for hype? in c/selfhosted@lemmy.world · 4d ago

• I’m not sure if Lepton improves on Waydroid for general apps as well, or just games. It would be great to see more weight behind Linux phones, but not sure if Valve intends to do it or not.

  on GrapheneOS Isn't Happy With Google Over Pixel's Widening Head Start in c/opensource@lemmy.ml · 4d ago

• But without SUVs, people would be embarrassed driving minivans and without huge trucks, there would be nothing to compensate for small penises.

  on Xiaomi Has a Luxury SUV for $31,000 That All Western Brands Should Fear in c/technology@lemmy.world · 4d ago

• A lot of us here hate streaming platforms because they’re designed to benefit greedy capitalists first and foremost. That doesn’t mean the core idea or the technology behind streaming platforms is bad, though. An open source project like Jellyfin that is designed purely to benefit the end user is something that legitimately enriches all of our lives.

  Similarly, just because scumbag companies like Anthropic and OpenAI are plundering the world for their own deranged benefit doesn’t make the underlying idea or technology bad. The tech is built on contributions from a lot of brilliant people over a long period of time whose names are not Sam Altman or Dario Amodei. Open LLMs that run on gaming rigs are quickly catching up with cloud-based models as well.

  At this point, I’m running Qwen 3.8 27B on a gaming GPU that has similar capabilities to Claude Sonnet 5 without giving Sam or Dario my data or $200 a month, just like I run Jellyfin without giving Netflix or Disney my money or data.

  on AI staff 'genuinely frightened' for humanity's future, ex-Anthropic researcher tells BBC in c/technology@lemmy.world · 5d ago

• Humanity should be way more frightened of greedy psychopaths running the world than any technology.

  on AI staff 'genuinely frightened' for humanity's future, ex-Anthropic researcher tells BBC in c/technology@lemmy.world · 5d ago

• I never wondered whether I was talking to anything approaching a divine being due to the number of mistakes. However, there have been points in the past with cloud models where I wondered whether I was actually talking to a human somewhere being paid a pittance to pretend to be AI. I self-host LLMs nowadays and am fairly confident there’s no human on the other end of OpenCode, but in the early days of cloud LLMs, I wouldn’t put it past them.

  on Cult of Anthropic Insiders Rumored to Be Worshipping Claude as a God in c/technology@lemmy.world · 6d ago

• Cool, although I’m assuming it requires pure CO₂ and would need to be combined with some other carbon capture mechanism.

  on Plasma and electricity convert CO₂ into methanol, butane and other useful chemicals in c/technology@lemmy.world · 6d ago

• LLMs improved significantly throughout 2024 until the first half 2025 and have now leveled off to more incremental gains instead of doubling benchmark scores within a few months. I think those days may be over. However, people who dismissed LLMs in 2024 and 2025 won’t have a correct understanding of their current capabilities. They’re useful and will probably continue with incremental improvements, but they’re also not a silver bullet that will magically produce high quality work without a competent human at the helm contributing most of the critical thinking and creativity.

  on Bernie Sanders says if AI is coming for workers, it should bring a four-day workweek with it in c/technology@lemmy.world · 17 Sep 2026

• Certainly possible that either a Trump executive order or congress will ban Chinese models in the United States. Legal at my company has already banned Chinese models to avoid the risk of the company relying on something that may become illegal. If Anthropic and OpenAI are successful in their marketing campign to paint Chinese models as dangerous and a threat to national security, they will have effectively kneecapped Chinese models in the U.S., at least.

  on AI Doomsaying Is an Aggressive Sales Pitch in c/technology@lemmy.world · 17 Sep 2026

• We passed the period of exponential LLM growth as of mid-2025 and now the gains are just incremental, which has given open models a chance to rapidly close the gap. Now the narrative is that the “dangerous” open LLMs need to get kneecapped while the “safer” closed LLMs continue with a healthy lead.

  on AI Doomsaying Is an Aggressive Sales Pitch in c/technology@lemmy.world · 17 Sep 2026

• Exactly. The Blender foundation has stated that it is a tool by humans and for humans, so no AI functionality is planned. However, there are still MCP extensions and the like. There’s nothing stopping anyone from using AI with Blender, but the Blender Foundation funding goes to core development. Not sure why Mozilla isn’t doing the same. If I really want a LLM controlling a browser, playwright-cli works fine anyway.

  on Mistral and Mozilla are bringing open, private and multilingual AI to your web browser: a partnership to bring privacy, control and choice to people using AI to browse online. in c/technology@lemmy.world · 17 Sep 2026

• They’re all trained on copyleft data, so they should by all rights be open source. Distilling them into open weight models is the next best thing, I suppose.

  on Y Combinator's Garry Tan wants US open-weight AI labs to 'distill' frontier models, too in c/technology@lemmy.world · 15 Sep 2026

• Curb it only on Thursdays?

  on EU to propose law curbing social media for children Thursday: EU chief Ursula von der Leyen is due to reveal an age limit for minors to access such platforms. in c/technology@lemmy.world · 15 Sep 2026

• I saw this article today summarizing a Mozilla report that open weight models are about 4 months behind the frontier: arstechnica.com/…/exclusive-open-chinese-models-c…

  Progress has slowed at the frontier and open weight is right on their heels. The general advice is that Fable and the like are best for niche use cases and to make an open weight model your daily driver because they’re almost as good and are a lot cheaper. In my case for personal use, I’m perfectly happy with Qwen 3.8 27B running locally and don’t care to deal with data collection or huge bills to get a bit stronger of a model.

  on AI’s recursive self-improvement might not come so quickly after all in c/technology@lemmy.world · 15 Sep 2026

• Looking at LLM benchmarks over time, there was still doubling of benchmark scores as of 2024 to mid-2025, then it inflected to more modest, incremental improvements, so sigmoid seems about right. Now we are seeing open models rapidly catching up now that the frontier has significantly slowed.

  on AI’s recursive self-improvement might not come so quickly after all in c/technology@lemmy.world · 15 Sep 2026

• The 0.4.0 release of llama.cpp just added the n-cpu-ffn flag for offloading weights for dense models to the CPU to complement the existing n-cpu-moe flag for MoE models. If you have a 16GB GPU and some extra RAM to spare, you might even be able to run Qwen 3.8 27B at a usable speed after some experimentation. I don’t own a 16GB card, so I haven’t tested it myself, and it might not work as well as I am imagining.

  on Yeah use AI.. luddite in c/programmer_humor@programming.dev · 15 Sep 2026

• I have a 7900 XTX with 24GB of VRAM where I am running Qwen 3.8 27B at 4 but quant with 200k context. I traded in my RTX 3070 and got it for around $700 recently. That model performs pretty similarly to Claude Sonnet 5 in performance for coding and certain other tasks and it “feels” similar to Sonnet 5 that I use as a daily driver at work. I was running Qwen 3.6 35B-A3B on the 3070 before and I still run it on my server with a 4060 and 32GB of RAM, and that model is more like Claude Haiku.

  I expect 3.8 35B-A3B should be released pretty soon and provide a nice bump in performance.

  I regularly follow this guy’s YouTube channel, and his latest video is about running a 177B model on his RTX 3060 via SSD offloading and getting Opus-like performance for certain tasks: m.youtube.com/watch?v=IH8XmxiwliQ.

  Supposedly, open weight models that run in a data center are like 4 months behind the frontier at this point, and what can run on a gaming PC is progressing rapidly as well while the exponential gains at the frontier have largely turned into incremental gains at this point.

  on Yeah use AI.. luddite in c/programmer_humor@programming.dev · 15 Sep 2026

• 

  on Sam Altman says OpenAI will not go public this year as AI fears grow in c/technology@lemmy.world · 15 Sep 2026

• I don’t see how they’re going to meet their revenue projections when open models are largely turning LLMs into a free commodity. So far in 2026, we’ve gone from local LLMs being a toy to Qwen 3.8 27B giving the latest Sonnet 5 a run for its money. The OpenCode desktop app is also almost at parity with the Claude desktop app as well and I’ve heard Pi is pretty good too.

  Now we are seeing techniques like a “phrase book” evolving to allowing a large portion of a huge model to reside on a SSD instead of in RAM / VRAM so that Opus-class models can run on a gaming PC. For example, this video is nuts: m.youtube.com/watch?v=IH8XmxiwliQ

  OpenAI and Anthropic know they’re cooked, so that’s why they’re talking all this nonsense trying to keep the shell game going a while longer. 🍿

  on Sam Altman says OpenAI will not go public this year as AI fears grow in c/technology@lemmy.world · 14 Sep 2026

• 3 bit and it performed like Opus 5 according to his tests, though likely not at that level on every task.

  on OpenAI boss and Elon Musk back calls to put brakes on ‘reckless’ AI development in c/technology@lemmy.world · 14 Sep 2026

• The one thing that hasn’t plateaued yet is the ability of open models running on a gaming PC to match OpenAI and Anthropic are doing. I just watched a video on the Codacus YouTube channel where the guy is running Qwen 3.8 Flash, a 177B model, on his RTX 3060.

  on OpenAI boss and Elon Musk back calls to put brakes on ‘reckless’ AI development in c/technology@lemmy.world · 13 Sep 2026

• I met my spouse over 10 years ago on a dating site, but now I believe they’ve all been bought up by one or two companies and have become completely enshittified. Dating sites would work if there were a Craigslist of dating sites.

  on Dating Apps Are Dying in c/technology@lemmy.world · 13 Sep 2026

• 

  on AI Didn't Steal the Mathematician's Work, Sam Altman Did in c/technology@lemmy.world · 12 Sep 2026

• Every time I see Dario, he reminds me of Pat from the 90s movie:

  

  on ‘The People Building AI Earnestly Believe That It Could Kill Us All’: Anthropic Researcher Quits Dramatically in c/technology@lemmy.world · 10 Sep 2026

• How long until smart TV manufacturers partner with AT&T, Verizon, and T-Mobile to collect your data even when you don’t give them a connection?

  on LG TV shown scanning LAN for third-party phones and other devices in c/technology@lemmy.world · 10 Sep 2026

• Works for now, but my fear is when that may not broadly be the case.

  on Your LG TV Might Be Spying on Your Entire Network While You Think It's Turned Off in c/technology@lemmy.world · 7 Sep 2026

• Agreed, although simply using LLMs to do your thinking for you is only going to decrease your value in the workplace and make you more replaceable. The goal needs to be learning how to use AI to expand your thinking, counteract your biases, improve the quality of your work, and accelerate your learning.

  on Linus Torvalds used AI to fix a Linux bug, and now LLM critics on Linux face an uncomfortable choice in c/technology@lemmy.world · 1 Sep 2026

• Frankly my dear, I don’t give a damn. Linux PCs will run the vast majority of games and the handful of exclusives that don’t run yet will run on an emulator at some point.

  on PlayStation Wants To Prove In Court That You Don't Own Your Digital Games in c/technology@lemmy.world · 1 Sep 2026

• Agreed, I was referring more to future updates. Obviously we’re good with what is available now.

  I can’t speak to pricing and availability outside the U.S., but it looks like the one I got went up $100:

  https://www.newegg.com/asrock-radeon-rx7900xtx-24g-radeon-rx-7900-xtx-24gb-graphics-card-triple-fans/p/N82E16814930084. 

  I traded in my 3070 and my final price was in the 700s. Last I looked, used ones were going for $800 on eBay vs. $1200 for a used 3090.

  I run 3.8 27B at q4 with q4 context up to 200k. Decode is generally in the 30s and pp starts in the 700s and drops to the 400s as context approaches 200k. I use mostly Sonnet 5 at work and I would rate this setup with the OpenCode desktop app as pretty comparable overall for coding at least. Let’s just say I have no reason to use any cloud models, not that I would do that voluntarily outside of being compelled to at work.

  on NVIDIA reportedly buys HuggingFace for 13 billion in c/technology@lemmy.world · 28 Aug 2026

• The core llama.cpp maintainers also work at HF and will now work for Nvidia I guess. Llama.cpp is a pretty significant part of the local LLM stack, especially since other tools like Ollama and LMStudio are just GUIs built on top.

  Local LLMs have gotten to the point where they are a serious threat to Anthropic and OpenAI, and Nvidia has a lot of skin in the game. If Nvidia wanted to do some serious damage to local LLMs, they are now in a position to do so.

  I’m also imagining they may try to squeeze out support for other GPU vendors. I’m using an AMD 7900 XTX to run Qwen 3.8 27B that I downloaded from HF to run on llama.cpp, which currently works like a dream. The 7900 XTX is the only sanely priced 24GB GPU left in 2026 (under $1k vs. $2k, $3k, $4k for Nvidia 24-32GB cards). Combined with OpenCode or Pi, a setup like this basically eliminates the need to use Anthropic or OpenAI products in the same way Jellyfin eliminates the need to use streaming services.

  I’m sure Nvidia and their buddies don’t like one thing I’ve said in this comment and may very well be plotting to put a stop to it, so the community may need to step up our game and get our eggs out of the big tech basket.

  on NVIDIA reportedly buys HuggingFace for 13 billion in c/technology@lemmy.world · 28 Aug 2026

• any actually useful real world tasks you need at least 24GB you can dedicate to the LLM alone, and if this is fully GPU vRAM

  Agreed. For anyone who doesn’t have a 24GB card, the AMD 7900 XTX is basically the only sanely priced 24GB GPU left in 2026 and can be had for under a grand new, whereas even a used 3090 is $1200 and $2000 new.

  Qwen 3.8 27B runs nicely on my XTX. I have to use Claude at work, and Qwen 3.8 in llama.cpp with the OpenCode desktop app leaves nothing to be desired in comparison. 3.8 27B is about on par with Sonnet 5 according to benchmarks, though I even prefer it over Opus because Qwen writes in clear, understandable language. We are at the point now where anyone with a 24GB GPU has little reason to use data center LLMs.

  on Nvidia has been in talks to acquire Hugging Face for more than $13 billion in c/technology@lemmy.world · 27 Aug 2026

• obtaining games through unauthorised means

  If games are going to have DRM, collect our data through unauthorized means, and be randomly taken away without our authorization, then exactly what social contract are we upholding?

  on Sony Reminds PS Users They Don't Own Digital Games in c/technology@lemmy.world · 23 Aug 2026

• Anthropic is aware, and it may be the fix here; the joke, not the verdict, may not survive that point.

  github.com/anthropics/claude-code/issues/77136

  on English ↔ Claudish Translator in c/programmer_humor@programming.dev · 23 Aug 2026

• Not that bootleg videos are really even watchable in the first place. Anyone who has ever misconfigured Radarr and got excited that a movie showed up earlier than expected, only to find out it’s a shitty hdcam release knows the disappointment.

  A screen recorded with Meta spy glasses would be worse. Hold your head still, damnit—no, look at the screen, not your popcorn, and chew quietly for fuck’s sake.

  on UK cinemas look at banning Meta smart glasses over piracy fears in c/technology@lemmy.world · 22 Aug 2026

• I suppose I’m not that up to date on Japanese government and politics. Can you elaborate?

  on Japan tried to build an operating system for the entire world, then the US government intervened in c/technology@lemmy.world · 21 Aug 2026

• Andrew Ng said artificial neural networks are like a “cartoon brain”, which I think is pretty apt. It’s cool that having billions of parameters (cartoon “synapses”) can result in emergent behavior where a machine can “reason”. That was the original idea from as far back as the 1940s, and now we have the compute and have found a transformer architecture that can achieve the original vision. Very cool indeed. What’s not cool is the tech bros turning the tech into a large scale game of blackjack.

  on LLMs and Humans are Cognitive Cousins – Sean Carroll podcast w/Dr. Chandra Sripada in c/technology@lemmy.world · 20 Aug 2026

• If streaming shitty shows doesn’t work out, they could always change the spelling to “Pee Cock” and do porn instead.

  on Peacock is raising prices across all of its streaming plans | TechCrunch in c/technology@lemmy.world · 19 Aug 2026

• It makes sense that if you can’t be bothered to write it, then I can’t be bothered to read it.

  Also, the term “AI slop” puts the blame on a machine instead of the lazy human who doesn’t give a shit and can’t be bothered to so much as proofread a machine’s output based on their half-assed input before subjecting other humans to it and wasting their time.

  on AI;DR (AI; Didn’t Read) in c/technology@lemmy.world · 18 Aug 2026

• I used to run Gitea and switched to Forgejo mainly because I prefer a community project over one run by a for-profit company. I never had any substantive complaints against Gitea during my time with it to be honest, but thought I’d get off of it before investing too much further in case it went in a bad direction.

  on GitHub Has an Availability Problem. Is It Time to Look Elsewhere? in c/selfhosted@lemmy.world · 17 Aug 2026

• Self-hosted Forgejo works well for my own use. My employer is moving to GitHub Enterprise where it’s a separate instance per company with certain infra shared. Should be a slight improvement, I guess, though probably not a bad idea to make sure internal tooling is built with the git forge being configurable. 😉

  on GitHub Has an Availability Problem. Is It Time to Look Elsewhere? in c/selfhosted@lemmy.world · 17 Aug 2026

• It unfortunately became a hammer for every nail. IMO, it’s a great tool when a SPA is actually merited.

  on You Should Stop Using Meta (Facebook) Products at All Costs in c/privacy@lemmy.ml · 16 Aug 2026

• I use React, but that’s about it.

  on You Should Stop Using Meta (Facebook) Products at All Costs in c/privacy@lemmy.ml · 16 Aug 2026

• The emergent behavior in huge models where it can “reason” instead of simply predicting the next token is fascinating. Artificial “neurons” built on statistics and linear algebra emerge to create something that legitimately has artificial intelligence. ANNs were conceived in the 1940s, building on centuries of development in statistical modeling and only now do we have the compute power to make this vision a reality.

  Yes, the “intelligence” has significant limitations and won’t be replacing human intelligence anytime soon, but it can actually be a useful tool if its limitations are kept in mind.

  The problem is of course the tech bros turning centuries of innovation they had no part in developing into a massive Ponzi scheme for their own profit.

  on How AI text watermarking works: a visual guide in c/technology@lemmy.world · 15 Aug 2026

• Same.

  on Firefox is now the last major browser that still supports uBlock Origin in c/technology@lemmy.world · 15 Aug 2026

• Right? Hunyuan 3D is open weight, free to use, and probably not too far off from the cloud-based services that charge per credit.

  on AI Generated 3D Models Flood Market, But Almost No One Is Buying Them in c/technology@lemmy.world · 15 Aug 2026