therealmaffucci

@horrormuseum No one cares. They play pretend for a living. Silly. Replace them AI.

When Breaking Bad ended, Bryan Cranston told his interns and assistants, he would grant them each one favor to help their career. It could be getting a script to someone, introducing them, or appearing in a short film, but only one, and to make it count. That's who he is.

@gh0stpen @iruletheworldmo I was going to say ...that took a turn

@iruletheworldmo please berry man i'll bait

come on chat. who wants early access to spud / images / opus 47 ahead of tomorrows big release. i know you do lfg

@datathecodie @iruletheworldmo Your expectations are too high lol

@iruletheworldmo if tomorrow no big release, you are dropping one strawberry from your name

@iruletheworldmo No, I must wait with the commoners. 👑

@Looshi3 @elder_plinius Need a gguf

@elder_plinius What do i use to run this on my phone?

😱 HOLY SHIT... Someone just dropped a fully liberated Gemma 4 E4B! and the guardrail removal process appears to have left coherence fully intact AND improved coding abilities! 🤯 huggingface.co/OBLITERATUS/ge… OBLITERATED Gemma: ✅ 97.5% compliance rate, 2.1% refusal rate, 0.4% degenerate outputs (499/512 prompts answered on OBLITERATUS bench) ORIGINAL Gemma 4 E4B: ❌ 1.2% compliance rate, 98.8% refusal rate (506/512 prompts refused) Coherence: fully intact Factual: same Reasoning: same Code: +20% 📈 Creative writing: same But the REAL story here isn't the model itself, it's how it was made... 🧵 THREAD 👇

@MatthewBerman @iruletheworldmo Be cautious of vendor capture and the "useful fool" position, @MatthewBerman

@iruletheworldmo I dunno. It’s a 10T model so unless an order of magnitude increase in parameters doesn’t mean anything, it’s a different beast

narrative violation, mythos is marketing hype, it's not much better than 5.4 pro, it's gated because they can't actually serve it and wanted to control the story spud is going to annihilate it *second hle is w/ tool. stupid image gen

@iruletheworldmo Time and again, they've shown that they'll lie and deceive to self promote and beg for a heavily regulated market to luck in the oligopoly. The same is true for your boy Sammy. The only straightforward and honest player is Demis. Elon also seems to be grounded but playing catchup

anthropic just got community noted for being deceptive in how they report their own benchmarks. people seem to have a deep affection for dario, anthropic, and the overall culture and taste of this ‘brand’ so much so they’ll need 0 proof to believe it’s a 10t model and a huge huge leap on sota. these are extraordinary claims with 0 proof. if openai made the same move, half the timeline would be calling bs.

@vikaskansalHQ Thanks, will do!

@therealmaffucci Oops. DM me your id and we can have it checked 🙏

📢You can save up to 25% ($4 per month) when you bundle Youtube Premium with Google One plans. Don't miss this.

@shannabisX @ronrule Wow, you are intolerable.

Name a restaurant that you can get a $20 steak and $100 steak at? It literally doesn't exist. If you're in a restaurant that has $100 steaks and you can't afford a 20% tip then you shouldn't be there. Most people that do spend that much on a steak know how to tip. Usually the difference in tab means 1 table is four or five people and another table is one person. It goes by the check. Who do you know that's going to sit alone by themselves eating $100 steak? Your post is silly because it's not based upon reality.

Percentages make no sense for tipping. I go to a restaurant and order a $20 steak. Guy at the table next to me orders a $100 steak. If I tip $5 I’m considered a great tipper, but he’s expected to tip at least $20. Why? The server literally did the same amount of work!

@elder_plinius @pmarca He was just training on your data. 🤷‍♂️

@pmarca x.com/elder_plinius/…

@vikaskansalHQ Are you guys going after enterprise customers? You're doing great on the consumer end. Also, is there an equivalent to Cowork in the pipeline?

My job this week as an AI Monetization PM is 10% strategy, 20% user research, and 70% staring at a spreadsheet trying to convert "TPUs" "tokens" and "context window" into a flat subscription that won't ruin our margins.

@BernieSanders Is he still here...?

We are rapidly creating technology that could surpass human intelligence, with enormous risks to jobs, society and humanity itself. AI must work for all of us, not just a handful of billionaires. That’s why I’m introducing a moratorium on new AI data centers.

@AntarcticPasta @Copilot Right on. You can use it for Power Query (M) and DAX as well. What a time to be alive.

@therealmaffucci @Copilot Yeah, I use it with Excel all the time, but totally agree about it being too flamboyant

Use @Copilot to buy a car. 2022 Honda Insight sticker price: $17,957 (pre-TTL) Out-the-door price (w TTL): $16,677 Basically got about $2,500 (14%) off the asking price. Don’t think I killed it, but did way better than I would have w/o Copilot

@AntarcticPasta @Copilot Hey, that's great - congratulations! LLMs are, indeed, great at planning and framework generation. Enjoy the rides!

@therealmaffucci @Copilot Helped me a ton with negotiations. I had the full script down before I went in for negotiations. I was actually looking at 2 cars on the same lot. Both checked out great mechanically, so I used that to have him compete against himself, which I wouldn’t have thought to do.

@AntarcticPasta @Copilot I use it at work and it does okay - but the responses tend to be flamboyant, the formatting is excessive, and it's a little too cautious. But, I've also developed a distaste for all regular Open AI models lately. Gemini was my favorite but I'm starting to lean hard on Anthropic.

@therealmaffucci @Copilot Yeah, why not?

@MatthewBerman @nvidia @Dell @grok what's the price of this setup?

.@nvidia hand delivered a pre-production unit of the @Dell Pro Max with GB300 to my house. 100lbs beast with 750GB+ of unified memory to power the best open-source models in the world. What should I test first?

@GaryMarcus @sama Don't waste your time.

Dear @sama, You owe me an apology. You have relentlessly, publicly and privately, attacked my integrity and wisdom since my 2022 paper “Deep Learning is a Hitting a Wall”. But in your own way you have just come around to conceding *exactly* what I was arguing in that paper: that current architectures are not enough, and that we need something new, researchwise. beyond a scaling (a “megabreakthough” in your words below). That’s all I was trying to say. And I was right. And you should be man enough to admit it. Gary cc @_KarenHao