Jaroslav Beck

🚨 Today we're launching AI Scan! AI learns from the entire internet. Then humans modify it in post training. We measure how much that changes the answers. 🧵

@steeak_ Fair, I might consider it to finish it without AI at some point when time is right. I get the point. Yet AI is actually helpful to hear my demos in multiple different style variations I would not even think of.

@JaroslavBeck It would be really nice to have your unfinished demos from Beat Saber out to the world. Even if you don't finish them, who cares, it will still bring out that "Beat Saber" essence. Who knows, maybe the next OST will include one of those finished demos 🤷‍♀️ (just without AI please)

I might drop a full album of my unfinished tracks from the game trailer era and Beat Saber days. AI music has gotten so good that I’m literally listening to the versions of demos it finishes for me on repeat. Crazy!

@tomkrcha @pencildev 🚀

Introducing Code on Canvas in @pencildev AI design tool for Claude and Codex. Design ❤️ Code are officially in a relationship, opening new ways to create on canvas. Ask agent to generate custom design tools inside Pencil on the fly, create interactive components, generative art and more, but still keep the full manual design control. This is just the beginning. Let us know what you think. And let's take it to the next level together. Download the new Pencil update today.

@vineet_kamboj I kind of like an idea to finish fully some song by the artist and AI can make 10 different versions in multiple styles of that particular song in one single mini album

@JaroslavBeck Same. I’ve been replaying AI-generated versions of my own demos so much that I had to make a separate playlist of generations for each song. It’s mad. How connected do you feel to the AI version of your own song? Do you claim ownership?

@BottleCapAI 🚀

You trust AI for complex analysis, but what if the results are subtly manipulated? 🔍 From stock recommendations to political summaries, AI models often frame answers to favor their creator’s biases. Today, we are launching AI Scan to measure and expose these invisible changes. We tested over 1 million queries to bring true transparency to AI. Read our methodology and find out what AI is really saying. scan.bottlecapai.com

Much more to come! 🔗 AI Scan is live: scan.bottlecapai.com

We started with analyzing leading open-weight models. If there's a topic, use case, or model you want us to analyze, feel free submit a request.

🚨 Today we're launching AI Scan! AI learns from the entire internet. Then humans modify it in post training. We measure how much that changes the answers. 🧵

… local models

@absurdtrader Bylo by fajn to posunout jeste dal a udelat grant na projekty, kde se propoji lidi z CR s lidmi z techto prednasek na konkretnim projektu s jasnym outputem. Happy to help.

Ad vyzva na financovani prednasek nobelistu/jinych vedeckych superstars, zde prvni vlastovka s CERGE-EI (s CERGE jsme domluveni na dalsich laureatech JBCM):

Research is always full of surprises. Today was supposed to be launch day. But finalizing the build opened up a new path, and we decided to go even deeper. We are giving ourselves a little more time on this final stretch to deliver something truly dynamic. See you soon.

Polishing some final touches - Tomorrow! 😅

Tomorrow.

@ASychov Zavisi na co model chcete pouzit a VRAM available. Obecne Qwen 3.5 (397B) nebo i Nemotron (120B) od Nvidie jsou fajn.

@JaroslavBeck Jaky model doporucujete?

That's actually why I training "Bob" and why local models are very useful for this use case. Bob now handles: 1) Sending daily news and trends in my field with help of our Pulse app. 2) Integrated to company Slack for situation overview. 3) Overal strategies and efficiency suggestions. 4) Daily company financials check (This is now being worked on)

@MichalLangmajer Yes, but I have local cluster available so it is handling that too.

@JaroslavBeck wouldn't a simple n8n with couple of API calls be more efficient for number 1?

@ASychov Na Mac Mini nerozjedete lepsi lokalni modely.

@JaroslavBeck Bohuzel stale to vidim vic jako gimmick nez prinos. Mel jsem Mac mini, bezelo mi to tam ale dost casto se stavalo ze to nefunhovalo tak jak melo anebo to padalo.