MattVidPro

Insane and over the top intros are back! So is the vibe check: Gemini 3.5 Flash feels pretty middle of the road. It's blazing fast, with mid intellect and a mid price point. It's worse for me than GPT 5.5 thinking (priced heavier) But 5.5 one shots prompts that 3.5 can't finish. youtube.com/watch?v=U1VMhx…

@mikek_la I think the pressure will make diamonds! New breeds of artificial intelligence will sprout 🌱 at least, that’s my hope!

@MattVidPro Intelligence is one thing, the current architecture/paradigm (context/RAG based) is another thing, and the hype cycle for that is basically over

@Wanderlosth I just constantly see opportunities to integrate AI. It is all in how you use the tool.

When you’re doing serious work, it’ll make you realize how useless LLMs are

@KaiXCreator Master architect

If AI builds everything for you, what’s your real job?

None of us, not ONE knows where the ceiling is on intelligence. Humans constantly scream into the void about intelligence yet we hardly know how to classify it, especially outside of ourselves, and even that’s not easy. Whatever is coming won’t be predictable

@Elaina43114880 My guess will be no 🤷🏻‍♂️ Google is struggling to provide the same intelligence at the same dollar. Jus my personal opinion.

Can Gemini 3.5 Pro beat GPT-5.6?

@dhruvkumar1805 @KaiXCreator Valve is a good example of a tech company with exceptional public pov

@KaiXCreator Valve. never releases games but nobody's mad about it somehow 💀

Name a tech company that literally nobody hates

@DesignCntrl For me if I’m using AI right it’s taking me to new places with plenty more work to do, but the kind of work I want to be doing. It will let me do more without wasting time on the little trivial bits.

@MattVidPro Technically if the Ai is doing all the work, the AI literate brains aren't doing much at all.

There is an incredible gap between those who are “literate” with AI models and those who are not. The top users are solving decades old unsolved math, creating custom software, genuinely impressive short films, the list goes on. Their brains are reaching farther than ever before.

Humanity went from throwing rocks (a pivotal advancement) to teaching them how to simulate thought & reason at a doctorate level. (also pivotal) We are not separate from what we invent. It will always be a part of us. Where would we be without rocks?!

@LexnLin Everything eventually gets recycled. Thats just nostalgia waiting to come back

AI has "killed": em-dash purple Inter, instrument_serif Sparkle icon 2*3grids ... all considered as slop now 😔

@mark_k They won’t even let us see unfiltered COT. There are other “developer tools” they possess internally to further steer and control models. AI saftey is tough to get right, but I would appreciate better steering tools and unfiltered cot.

@MattVidPro It's too powerful for us to play with, the AI could end up saying naughty words! 😱

None of the big AI chatbots let you edit the assistant response. You can only edit your own message. Technically, allowing both would be trivial. In the prompt, it’s all just text. The only real reason they don’t allow it is, of course, MUH SAFETY. You could steer the conversation in ways the lab doesn’t want to see.

Grok foundation model V9-Medium (1.5T) has finished training. Evals look good. A lot of Cursor data was added in supplementary training and there is more to come. Fine-tuning is underway and reinforcement learning begins in a few days. 2 to 3 weeks to public release. This will be a major improvement over the 0.5T v8-small that currently serves all Grok production traffic, especially for difficult coding tasks.

Open AI’s “refocusing” has done them wonders. I speculate they’re ahead of the game internally optimizing Quality, Speed, & Cost. all of these companies have models they can’t yet serve. Everyone has the same problem: deliver upfront value while bolstering infrastructure. $$$$

Seedance 2.0 on release was so far ahead. Idk what they did, but it worked. Really the first model that can create believable enough consistent scenes in order to support a narrative. It still comes down to the person who wields the magic tool. But that gap will close.

We now know that with an appropriate harness both Mythos and GPT-5.5 can reproduce what our internal model did in one-shot for the unit distance problem. Clearly there is an insane overhang of capabilities with this generation of models, and no ceiling in sight for what scientific advances they can bring. You can go and try to discover new things with 5.5 right now!

Insane and over the top intros are back! So is the vibe check: Gemini 3.5 Flash feels pretty middle of the road. It's blazing fast, with mid intellect and a mid price point. It's worse for me than GPT 5.5 thinking (priced heavier) But 5.5 one shots prompts that 3.5 can't finish. youtube.com/watch?v=U1VMhx…

Today we are launching Music v2. Better vocals, instrumentation, and arrangement across every genre, improved multilingual support plus capabilities that weren't possible before.

Erdős problem #90 has been open for decades. Over the weekend a mathematician tested whether Claude Mythos could solve it. It did. But what caught my attention: Mythos didn't replicate the known approach from OpenAI's #1196 solution. It repeatedly settled on a different argument, one the mathematician called cleaner, with "no analytic complications." Air-gapped, no internet, no information leakage. GPT-5.5 solved numerous Erdős problems earlier this year. DeepMind's Nexus knocked out 9. Now Mythos, with a cleaner proof than the one that already existed. Problems that survived 80 years are falling in weeks.