β¨AI should be about empowering humans, building understanding, and making dreams realities. π©βπ» DevRel Eng. Lead @GoogleDeepMind ex-@GitHub || views = my own!
179Β°56β²39.4β³, +0Β°2β²46.2β³ 8k pc
- Tweets 71,019
- Following 1,948
- Followers 76,255
- Likes 119,548
Pinned Tweet
To get access to the latest Gemini models, including the most recent experimental release that's currently at the top of the @lmsysorg leaderboard, use @googleaistudio.
You can also turn on features like code execution; use extended context (2M+); and turn off safety settings:
19
55
398
217,319
me: i don't really do romance
two enemy agents writing letters across collapsing timelines:
4
29
2,473
just started reading "This is how you lose the time war" with zero context, and every other paragraph feels like an allusion to AI or agents π
"i have read the right books / to interpret your looks"
11
2,333
"AI can never stand in genuine wonder before the beauty of God's creation, and it can never replace the unique gift that you are to the world."
βAI can process information quickly, but it cannot replace human intelligence. And donβt ask it to do your homework for you.β β Pope Leo XIV
7
8
134
7,114
π There's a new @Lucasfilm Skywalker Studios general store near where I live in the Presidio!
5
1
45
3,383
π©βπ» Paige Bailey retweeted
Nail on the head. My calibration is completely cooked because building something that would have felt like fucking witchcraft 18 months ago now becomes infuriating when it takes 20 minutes. The capability inflation is so unceasing that we devalue it almost immediately.
thereβs so much confusion about whatβs happening in AI because if you donβt use coding models every day itβs hard to understand how quickly progress is accelerating but if you do, you likely have AI psychosis to a degree that itβs impossible to explain it in an understandable way
1
1
9
1,959
agree. a model is a mirror made out of all of us -- but someone still has to aim it, and to decide what's worth reflecting
In this era of artificial intelligence, it is becoming urgent to distinguish human art from what machines produce. There is an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation based on millions of images created by others. Algorithms lack the spark of humanity. For this reason, the Church wishes to renew an alliance with artists and cultural institutions to safeguard our humanity.
3
2
43
2,502
π₯π©πΌβπ» Retro computing and hacker vibes at @modal's Runtime conference (aka "infrastructure Coachella") are unparalleled.
Also really loved that they have the JEV team speaking this afternoon! Stacked company list - @sarahookr @nvidia @AnthropicAI @swyx @marksaroufim and more.
3
2
46
2,911
humans: small. squishy. cannot fold fitted sheets. need snacks every three minutes
also humans: figure out how to throw computers into space, rovers on mars, a car in orbit, and (as of this week!) tpus, too
we are so weird and i love us, here's to decades more of looking up and asking "what if we did things *over there*"
3
3
74
3,505
π©βπ» Paige Bailey retweeted
The fundamental tension w AI and education is that I think schools ought to (a) teach people how to think and (b) prepare them for the workforce (or, at the very least, not actively harm their employability).
Those goals push in opposite directions.
AI interferes with thinking but itβs also baked into the work of software, finance, legal, marketing, etc. If you use too much AI in college, youβll be dumb; never use it for four years, and youβll be in a disadvantaged position for the most popular high-paying entry level jobs.
Something like this below seems like at least a thoughtful way of addressing that tension.
π§΅UChicago is leading the way on AI in higher education.
1. All students will take an analog writing class.
2. Core classes in the humanities and social sciences will prohibit technology in class.
3. A new requirement in computational thinking will likely be added to the core.
47
57
432
65,329
πΈ Gemini 4 Argon is currently #1 on @ValsAI Finance Agent v2 leaderboard: vals.ai/benchmarks/fabv2
9
3
114
4,661
π The world desperately needs more long-horizon tasks for models to tackle! Congrats to the @Android team (@matthewmccull et al!) for getting this released.
Android Bench 2.0 expands beyond bug fixes and adding tiny features into complex development tasks that engineers typically spend several days to a week on. The eval slates in at ~30 long-term tasks, including creating new apps, migrating dependencies and designs, and porting cross-platform apps to Android.
π’ Introducing Android Bench 2.0.
We've leveled up our AI evaluation framework to handle real-world, multi-day engineering challengesβfrom building apps from scratch to migrating cross-platform codebases to Android.
Letβs dive into whatβs new. π§΅ππ½
ALT Android Bench 2.0 dashboard with category filter buttons.
3
3
45
4,276
The best model for a task is the one that can do it quickly, reliably well, and at a budget-friendly price. π°
π¨ Gemini 4 Argon is ridiculously cheap too
It's between GPT-6 Astra and GPT-6.1 Sol in terms of cost, while being Astra level in intelligence
Google absolutely nailed the price/performance
16
3
141
7,196
Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #8 in Code Arena: WebDev with 1679 pts!
This release has reshaped the Text Arena Pareto frontier with a blended $8/MToken! Gemini 4 Argon (High) is now the most cost efficient model, see its placement on Pareto frontier below.
In the Text Arena, Gemini 4 Argon (High) ranks #1 in Coding, Hard Prompts, Instruction Following, Longer Query, and Creative Writing. It also leads every occupational domain evaluated, with additional #1 spots in English, Non-English, Chinese, and Russian.
This model is +20 points above the #2 ranked Claude Opus 4.6 (High), and a huge leap from Googleβs previous release, Gemini 3.8 Flash (High) at #11!
In Code Arena: WebDev, Gemini 4 Argon (High) gained +96 points from Gemini 3.8 Flash (High), and went from #29 to #8.
Congrats to the @GoogleDeepMind team on this impressive frontier release!
Introducing Gemini 4 Argon β our new frontier model.
Itβs built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense β rolling out today to a set of trusted testers through our Fairwind Program.
135
320
3,611
816,839
