Reposted by Cool Edpriscilla page @priscillapage.bsky.social · 03/10/2026and like, his entire family busts in with a cake to wish this fully nude man in the bathtub a happy birthday. his kids goof off with him a little, chit chat. legit bonkers stuff. 1383855
Cool Ed @eddasaurus.bsky.social · 30/09/2026This year I just used Gemini Pro, asked a few questions to get started. Oh is it too late in the year? “Suboptimal but you should be fine.” Dive in. Take a pic, am I doing this right? “Cut it a little further down and clean it up” Oh here’s something weird what do I do? Etc etc. Invaluable assist 010
Cool Ed @eddasaurus.bsky.social · 30/09/2026For example, I’ve struggled with keeping some bushes under control because you’re supposed to prune them in a particular way at a certain time of year. Spring rolls around, oh I should trim them, let me remind myself how, find conflicting info so check multiple sources, end up on specialty Reddit, ☠️ 110
Cool Ed @eddasaurus.bsky.social · 30/09/2026At least in my experience. Not formally ADHD but sometimes I wonder, and I definitely deal with tons of inertia trying to learn *everything* before doing a thing. And usually taking so long to do that my window for doing it closes, or knowing this will happen dissuades me from even starting 111
Cool Ed @eddasaurus.bsky.social · 30/09/2026The ADHD individual is Popeye and the LLM is the spinach, giving them a hyper concentrated dose of targeted information sufficient to get them through the thing before they get distracted and go down a rabbit hole 120
Cool Ed @eddasaurus.bsky.social · 28/09/2026This is actually an all timer “so close to getting it post”. The thing you acknowledge could lead to reasoning happened. Two years ago. With the minor caveat that it’s the model talking to itself. And that’s directly responsible for the very achievements you’re skeptical of. Incredible 120
Cool Ed @eddasaurus.bsky.social · 28/09/2026It’s not really secret sauce, and it’s not just Claude. AI vendors have been developing “reasoning” models since late 2024, and they do the “interactions between agents” internally. Break down problem, research data as needed, pursue multiple approaches, critique and revise candidate responses 141
Cool Ed @eddasaurus.bsky.social · 25/09/2026Great point. That show was amazing and I rarely see anyone discuss it, nice to see that it’s not completely forgotten 000
Cool Ed @eddasaurus.bsky.social · 25/09/2026No Venture Bros? The Expanse?? List invalid. Band of Brothers out of the top 10 is pretty egregious too. 100
Cool Ed @eddasaurus.bsky.social · 24/09/2026Oh this feels relevant to your broader point, but I don’t know how to put it into practice. Maybe it’s related to the helpful bot idea you mentioned, or maybe it’s as simple as sharing success stories more. bsky.app/profile/edda... 130
Cool Ed @eddasaurus.bsky.social · 22/09/2026🎶 Resettin’ all the limits 🎶static.klipy.comScarface: Tony Montana on Phone & Money CountingAlt: Scarface: Tony Montana on Phone & Money Counting GIF of the famous montage scene set to Push it to the Limit 020
Cool Ed @eddasaurus.bsky.social · 22/09/2026static.klipy.comZandig John ZandigAlt: Zandig saying JEEZUS! 010
Cool Ed @eddasaurus.bsky.social · 20/09/2026I’d argue that there’s value in teaching how to evaluate what comes out of AI than how you put info into it. How to verify information, how to sense what *needs* verification and what you can take at face value, how to avoid confirmation bias, etc. Critical thinking is key to maximizing these tools 150
Cool Ed @eddasaurus.bsky.social · 20/09/2026static.klipy.comSimpsons: He's History's Greatest Monster!Alt: Simpsons: He's History's Greatest Monster! 031
Cool Ed @eddasaurus.bsky.social · 19/09/2026This article shows Grok 4’s RL compute equaling or exceeding pretraining last summer www.lesswrong.com/posts/xpj6Kh...lesswrong.comHow Well Does RL Scale? — LessWrongSummary: RL-training for LLMs scales surprisingly poorly. Most of its gains are from allowing LLMs to productively use longer chains of thought, allo… 160
Cool Ed @eddasaurus.bsky.social · 18/09/2026Yes. And this paradigm shift started in September 2024. Lots of people still arguing based on a *very* outdated mental model 1160
Cool Ed @eddasaurus.bsky.social · 18/09/2026Which specific month does the divergence begin? Was it October? 000
Cool Ed @eddasaurus.bsky.social · 18/09/2026interactions limited because I’m not interested in hearing “tHaTs ThE wRoNg KiNd oF LeFt” 000
Cool Ed @eddasaurus.bsky.social · 18/09/2026I’m old enough to remember an online left that believed copyright is theft, information wants to be free, no one owns the commons, and the internet should be open. Pretty sure the consistent stance is LLMs are fine, they should be open source, and everyone should have free access to SOTA models 1274
Cool Ed @eddasaurus.bsky.social · 16/09/2026Some are. But the mental model of “statistical model that predicts what comes next” is no longer usefully descriptive when the response is the sum total of what I described rather than just the prompt you entered into the chat window. Particularly when it runs in a loop until it gets a good answer 200
Cool Ed @eddasaurus.bsky.social · 16/09/2026Layers that intercept the prompt and do things with it. Things like deconstruct the prompt to see what data needs to be fetched or external tools need calling, generate a detailed plan of attack, farm off work to subsystem LLMs, combine the results of it all to generate a candidate answer and review 200
Cool Ed @eddasaurus.bsky.social · 16/09/2026Your description has been incomplete and outdated for over two years now. Tool use, integrated search, reasoning models, multi-modal, agent swarms, etc. The statistical model you describe is still in there but there are many layers on top of it 160
Reposted by Cool EdMicah @rincewind.run · 15/09/2026wake up, get online, the head of the fbi is defending bestiality on the floor of the senate that’s enough internet for the day already 22715139
Cool Ed @eddasaurus.bsky.social · 13/09/2026Not the main point but what the fuck is up with this picture? What is it trying to convey? Is the crow mixed into the mystery meat? Is the crow the one that put all those broken eggs on top of the ground beef? What’s with the diced vegetables? World’s shittiest steak tartare??? 000
Cool Ed @eddasaurus.bsky.social · 08/09/2026Claude, invent me a room temperature super conductor. Make no mistakes 151
Cool Ed @eddasaurus.bsky.social · 05/09/2026rice that’s fried is justified add a protein and a chosen side 0325
Cool Ed @eddasaurus.bsky.social · 03/09/2026I wish Deadlock still covered AEW, they had a bit where they’d riff on Tony intentionally redoing angles that previously featured people who’ve since left the company youtu.be/GFi7HznP-PY?...youtu.beTONY KHAN IMPRESSIONS! | DEADLOCK COMPILATIONSYouTube video by Lost The Camera 020
Cool Ed @eddasaurus.bsky.social · 28/08/2026That’s a ceiling, not a projection. Total addressable market is basically the entire universe that they could potentially sell into, which is important for investors to know how much room to grow there *could* be. Challenges to capture that market, competition, etc, are all implicit and expected 030
Cool Ed @eddasaurus.bsky.social · 25/08/2026It’s not just higher demand, it’s also more varied. If I’m an enterprise client serving 4o for an internal chatbot I built maybe I don’t need the latest tier of GPUs. And perhaps more importantly, speed and failure rate are not huge problems with async RL runs to improve reasoning functionality 040
Cool Ed @eddasaurus.bsky.social · 24/08/2026Dead from ass cancer at 32? Karma is definitely real. 022
Cool Ed @eddasaurus.bsky.social · 23/08/2026Also their various Stone Cold impressions! youtu.be/ZEzU6MDSwpA?...youtu.be[ᴅᴇᴀᴅʟᴏᴄᴋ] James and Jawnnys excellent Steve Austin/JR seg from Podcast 346 #deadlockpodcastYouTube video by 𝙥𝙖𝙮𝙣𝙚𝙭𝙠𝙞𝙡𝙡𝙚𝙧 𝙧𝙚𝙫𝙞𝙚𝙬𝙨 110
Reposted by Cool EdJon Bois @jonbois.bsky.social · 22/08/2026i get it but this song must be played at 400% volume. sorry 1344819
Cool Ed @eddasaurus.bsky.social · 15/08/2026Playing this song every time his bullshit is covered in the AEW documentary youtu.be/O0bk6Hbz0ZU?...youtu.beThe Order of DeathYouTube video by Public Image Limited - Topic 010
