Alex Astrum
@alexastrum
Followers
777
Following
215
Media
60
Statuses
1K
google, dontdie, and prometheus rising 🔺
Joined April 2009
Introducing our new SOTA Gemini 2.5 Computer Use model, trained to take 13 different actions and navigate a browser. This is just the first step in the Gemini computer use story : )
113
180
2K
Gemini CLI now has an official catalog of Extension that will make using and extending it for your own need so much simpler. https://t.co/wRDVfRApgP
blog.google
Make Gemini CLI uniquely yours by connecting it to your everyday workflows and the tools you use most.
0
0
1
thanks the folks that attended live or viewed the livestream
Firebase Studio: AI-powered JavaScript development Build features faster—without cutting corners. Meet Firebase Studio, Google’s AI that scaffolds your app with security, data, and deploys baked in. What you’ll learn ⚡ Generate components, APIs, auth, and data models with
0
0
1
it was a good week for open source ai.
🚨 New Top Open Model Update! A relative newcomer to the Arena, @zai_org's GLM-4.6 takes the clear, undisputed #1 spot for Top Open Model. 🏆 It also ranks #4 overall, which is not an easy feat! The next top open model, DeepSeek R1 0528, has been the standing champion for
0
0
0
after trying the Sora app over the weekend, I have to admit that a lot of things are just right for it to take on popular platforms like tiktok. unlike the popular influencer @theo opinion https://t.co/khH7X9CPrY, i'm seeing potential for RL spirals of virality. some signs of
1
0
0
There's still time to register for the upcoming Voice AI + @googleaistudio + @pipecat_ai @trydaily hackathon this weekend! October 11th // @ycombinator HQ
Voice and multimodal AI Hackathon at @ycombinator on October 11th. Engineers from @trydaily and the @pipecat_ai core team will be there. Join us to build something new, learn from other fun people, and win prizes like a guaranteed Y Combinator interview!
6
11
48
AI standup comedy might just be the next meta
@RocketsTTV @pitdesi @rob_mcrobberson Here are a few "equally innovative" services you could offer affluent Bay Area folks for easy cash: - $25k personalized emoji consultation for your Twitter bio (includes aura matching). - $20k artisanal air selection: I sample and recommend the best oxygen blends for your home
0
0
0
This is insane! The new Gemini Flash model released yesterday has the same accuracy as o3, but it is 2x faster and 4x cheaper for browser agent tasks. I ran evaluations the whole day and could not believe this. The previous gemini-2.5-flash had only 71% on this benchmark.
Introducing two new Gemini 2.5 models (Flash and Flash-Lite) which are more intelligent, cost effective, and token efficient. You can keep up with our latest models through `gemini-flash-latest` and `gemini-flash-lite-latest`!!
58
169
2K
🚨 DeepMind finally dropped the Veo3 paper which shows what we all realize from playing with video-gen models. Just like LLMs, visual reasoning on is an emergent property of training on tons of video. It can solve tasks not explicitly in training data. "Veo 3 is the GPT-3
63
322
3K
2026 might be all about embodied ai. Till we get there, here is something you can use today paired with your favourite DIY Raspberry Pi robot.
It is very clear that robotics is the future, so seeing both the capability transfer from mainline Gemini to this use case but also the pace of research progress specific to embodied reasoning is amazing! https://t.co/fiQdM0n6SZ
0
0
0
Introducing two new Gemini 2.5 models (Flash and Flash-Lite) which are more intelligent, cost effective, and token efficient. You can keep up with our latest models through `gemini-flash-latest` and `gemini-flash-lite-latest`!!
169
224
3K
Tiny models are great for educational, practical content. This one is also nostalgically beautiful. Read through, try it out. The best way to understand how technology works is to build it yourself at a tiny scale
I spent the past month reimplementing DeepMind’s Genie 3 world model from scratch Ended up making TinyWorlds, a 3M parameter world model capable of generating playable game environments demo below + everything I learned in thread (full repo at the end)👇🏼
1
0
6
This is just the start of the 2nd act for the Live API, so much more to come over the next few months, the frontier is omnipresent AI agents you interact with by voice and camera! Learn more:
19
16
309
good essay on the chronology of a weird emerging symbiosis. the material is somewhat heavy, and the author expresses obvious bias against some belief systems and neurodiversity traits, but the analysis is otherwise solid. https://t.co/0vMxWiiYM7
0
0
0
nice model based on qwen3 2b. solid performance when used in well lit scenarios with clearly defined objects being inquired. will hallucinate when used with images missing required objects, but containing other semantically close objects (eg. hallucinate and tag fruits, if the
1/ Introducing Isaac 0.1 — our first perceptive-language model. 2B params, open weights. Matches or beats models significantly larger on core perception. We are pushing the efficient frontier for physical AI. https://t.co/dJ1Wjh2ARK
0
0
0
given that 1/2 population is toddlers and other folks not consuming internet media, 50 vids/person is highly disturbing, espcially since tiktok & insta is not even included here.
this guy felt comfortable standing in front of a screen that said this. this is how stupid all of you are. dumb, idiotic, pea for brains cattle that doesn't even know what is happening to them.
0
0
2
Can AI help understand how the brain learns to see the world? Our latest study, led by @JRaugel from FAIR at @AIatMeta and @ENS_ULM, is now out! 📄 https://t.co/y2Y3GP3bI5 🧵 A thread:
30
305
2K
Less than 48 hours ago, xAI dropped Grok Code Fast 1. And it is already breaking brains. The speed is unreal. And it is FREE until September 10th. 10 wild examples + how to try: 1. Building battle simulator prototype https://t.co/p8vkOkBQdW
35
70
575
Even respectful folks don’t read their ai generated newsletters anymore. Seeing full paragraphs of ai slop (like ai follow up propositions) is disappointing. I won’t give names. The consumer market content creation craft is dead. Long live hybrid intelligence!
0
0
0