Zetaphor @zetaphor.com · 14/09/2026I've been using this, it's the same UX as the VSCode bottom terminal which is nice github.com/siberiah2o/d...github.comGitHub - siberiah2o/dsh-plugin-terminal: Bottom terminal panel plugin for DeepSeek Harness (DSH Web GUI)Bottom terminal panel plugin for DeepSeek Harness (DSH Web GUI) - siberiah2o/dsh-plugin-terminal 110
Zetaphor @zetaphor.com · 13/09/2026pi.dev is the goat but requires you to bring/roll your own batteries I've recently switched to Deepseek Harness as it's even *more* extensible and has a web UI and subagents by default 110
Zetaphor @zetaphor.com · 13/09/2026Okay unironically I just got DSH pilled last night and I'm very excited and want to find other people doing stuff with it. The discord link is dead 😞 000
Reposted by ZetaphorRyan @depatchedmode.bsky.social · 12/09/2026We have officially left the “hand transcribed by monks” era of software. 628142
Zetaphor @zetaphor.com · 12/09/2026Now that I have 3.8FN running at a regular 20-40 t/s it's completely replaced proprietary models for me. It is now the quality and speed I would expect from Opus 4.6, which has been my 'ol reliable since it launched 010
Zetaphor @zetaphor.com · 20/08/2026It just occurred to me that the discontinuation of the penny as currency means eventually there will be no more ass pennies in circulation 😞 000
Zetaphor @zetaphor.com · 07/07/2026Throw the entire stack into a docker-compose file and your entire automated media piracy and playback network is contained in a text file you can toss on any server with an internet connection. I use an old mini PC to run mine 000
Zetaphor @zetaphor.com · 07/07/2026Radarr and Sonarr are amazing. Throw in Bazarr for subtitles. Sabnzbd if you're willing to pay for usenet, otherwise Proawlarr and Jackett if you torrent. I pay three different usenet providers and it's still less money/time/sources than if I were try and get the same content through official means 130
Zetaphor @zetaphor.com · 25/06/2026Who even is this dude and why is his whining relevant. He'll be dead by the time the shit kicks into even higher gear 100
Reposted by Zetaphorjoe bodnar @joeabodnar.bsky.social · 11/06/2026Bummed by this OpenAI report. It's Spamouflage. It's not new and it's not impactful. If I was cynical, which I'm not, I'd say this report was just meant to stir up anti-China sentiment. openai.com/index/prc-li...openai.comPRC-linked influence operations are targeting AI debates in the USA new report from OpenAI details PRC-linked influence operations using AI to target U.S. tech debates, data center narratives, tariffs, and false claims about ChatGPT. 233
Zetaphor @zetaphor.com · 11/06/2026My bullshit posts on Bluesky are more influential than this campaign, but they felt this warranted a 14 page report with the PRC prominently called out in the title 000
Zetaphor @zetaphor.com · 11/06/2026The line immediately after that speaks volumes about their motives: "Nevertheless, this operation targeting OpenAI, a private company operating in a strategically important industry, should raise concerns." 100
Zetaphor @zetaphor.com · 11/06/2026OpenAI: OMG THIS IS DANGEROUS!!! Also OpenAI: "Most of the social media posts we identified generated little or no observable engagement. We found no evidence that the false claims about ChatGPT user data being compromised were amplified by authentic high-reach accounts or beyond the X platform." 100
Zetaphor @zetaphor.com · 11/06/2026They see the rising tide coming up around them, brought on by open weight models from mostly Chinese labs. Why do we need $500bn for Stargate when more and more of our needs can be met by a 3090? This post is 3 years old and aging like a fine wine newsletter.semianalysis.com/p/google-we-...newsletter.semianalysis.comGoogle "We Have No Moat, And Neither Does OpenAI"Leaked Internal Google Document Claims Open Source AI Will Outcompete Google and OpenAI 000
Zetaphor @zetaphor.com · 11/06/2026I haven't been able to shake this since reading it a couple weeks ago qntm.org/mmacevedo I haven't seen anyone other than Anthropic seriously propose the idea of being concerned about a models wellbeing and state of mind. I shudder to think what will be done as local AI increases similitudeqntm.orgLenaA note from the author You can now buy this story as part of my collection, Valuable Humans in Transit and Other Stories. This collection also includes a sequel story, titled "Driver". This articl... 000
Zetaphor @zetaphor.com · 11/06/2026I may not share your tastes but I will fight to the death for your right to use local models for whatever degeneracy you choose to get up to ✊ 040
Zetaphor @zetaphor.com · 11/06/2026That's what I get for coming in from a feed and not reading the room lmao 130
Zetaphor @zetaphor.com · 11/06/2026Any specific use case for the uncensored models or just prefer to not have to deal with guardrails? I'm always worried about diminished quality with those variants 210
Zetaphor @zetaphor.com · 11/06/2026wtf now I can reply? I thought it was strange that I got the reply dialog but the actual reply button was disabled. Now it seems to work fine? I swear the button remained disabled when I tried before. Either way point stands, don't be a dick on the internet 000
Zetaphor @zetaphor.com · 11/06/2026Quote posting since this person limits replies. I understand why, though it's always a bummer when you feel like you have something useful to contribute to the conversation. Being a shitty person online is like pissing in the pool, you're making things worse for everyone 😔 110
Zetaphor @zetaphor.com · 11/06/2026I do this regularly looking for repeat failure patterns sessions to surface as either new/refined instructions for my global agents.md or for new Pi extensions. A recent example was the model repeatedly trying to run sudo, so I now have an extension that blocks it and prompts to ask me instead 200
Zetaphor @zetaphor.com · 11/06/2026It looks like Microsoft and Apple are getting serious about pushing more compute to the edge, their motives are certainly questionable (especially in the case of Microsoft), but I prefer to see it as a silver lining that we can leverage to our advantage 010
Zetaphor @zetaphor.com · 11/06/2026As a Linux user I sadly have to agree. The story of local inference is still pretty rough. Software like LM Studio/Msty/Jan are helping but there's still a lot to navigate wrt to the headache that is CUDA/ROCm/MLX/Vulkan runtimes. One hope is the AI bubble produces better solutions before popping 110
Zetaphor @zetaphor.com · 11/06/2026A perfect example is this thread. My original attempt was far beyond the character limits and not well structured for a thread layout as it was stream of conscious style of writing. Threw it over to a local model to help me brainstorm a less wordy version that would break down well into two posts 010
Zetaphor @zetaphor.com · 11/06/2026Between SSMs and quantization, PhD-level AI is hitting commodity hardware faster than consumer needs are rising. We’ll reach a point where local models are like mobile gaming, maybe not the SOTA high-end PC equivalent, but more than enough for the average person’s daily needs. (2/2) 210
Zetaphor @zetaphor.com · 11/06/2026Great thread, but I’d argue Jevon’s Paradox isn’t absolute. Some tech has a utility ceiling. Take PCs: after the 1GHz threshold, "basic" systems became good enough for most. Faster speeds became a luxury, not a necessity. I think we're hitting that same "good enough" threshold for LLMs. (1/2) 210
Zetaphor @zetaphor.com · 10/06/2026Ignore this and take the advice from the humans in your replies, these models have a tendency to suggest outdated solutions 000
Zetaphor @zetaphor.com · 09/06/2026Curious how paddle would do for this use case github.com/PADDLEPADDLE...github.comGitHub - PaddlePaddle/PaddleOCR: Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ l...Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages. - PaddlePaddle/Paddl... 000
Zetaphor @zetaphor.com · 09/06/2026Please consider something other than ollama sleepingrobots.com/dreams/stop-...sleepingrobots.comFriends Don't Let Friends Use Ollama | Sleeping RobotsOllama gained traction by being the first easy llama.cpp wrapper, then spent years dodging attribution, misleading users, and pivoting to cloud, all while riding VC money earned on someone else's engi... 000
Zetaphor @zetaphor.com · 05/06/2026They're doing atproto shenanigans and sending replies to non-existent posts in an attempt to funnel you into a preorder 030
Zetaphor @zetaphor.com · 05/06/2026For anyone getting replies or seeing posts from this or the similar @clawphones.bsky.social, these are both very obvious scams trying to cash in on the branding of @nousresearch.com Avoid like the plague #locallama for visibility 121
Zetaphor @zetaphor.com · 31/05/2026😳 Tell me you have the Noctua fan and also installed the 80mm front fan? 110
Reposted by ZetaphorAstra ⎔ @astrra.space · 31/05/2026oh my fucking god we were _this_ close to having a normal common way to distribute Linux software 1113310
Zetaphor @zetaphor.com · 28/05/2026You're taking on a MASSIVE legal liability! I would have stuck to only allowing BYOS 000
