Last night I had to ask my wife who she was spying for.
As I realized there is no way I could have landed someone so amazing without motives...
She replied instantly that she married me for my money... (dont know what bank account she was looking at)
Unfair that CEOs who aren’t into Asian girls get fewer dating options from spy agencies.
Where are the French/Russian/Iranian honeypots? Not one spy girl sent my way. Are they even trying?
France is excused — their decision-makers aren’t into girls. Everyone else: try harder!
2
31
AI builders on a budget,
After running a test do you ask your agent how you can reduce the cost even more?
I just did on auto pilot and here were the costs...
4
The is really good, Gen AI UI is some of the worst if not given good direction.
Take 5 minutes to skim if you are a builder without time and frustrated at UI
Today we are releasing A2UI 1.0, now faster, easier, and stable, an open protocol and framework for generative UI across web, Android, iOS, wearables, TVs, and any other surface.
👇 Check out what shipped
11
This week is Christmas for AI peeps all around, from Meta to the big G.
Antigravity, decision 1 toos, voice transcription from Microsoft models, Ember, Carbon, and even many smaller company launches!
Make it a long weekend and enjoy
20
Tech PSA announcement of the year
Look at your PC mouse. That's roughly how big your kidney is.
You actually have two. You can survive with one, which explains the classic "sell a kidney to buy an RTX 5090" joke. 😂
In tech terms, your body comes with RAID 1 enabled by default. Pretty good hardware design, right?
But here's the thing.
If you're coding until 2 AM, living on energy drinks, barely drinking water, and taking ibuprofen every time your head hurts, don't assume that second kidney is just sitting there as a spare.
Your kidneys aren't replaceable hardware.
Anyway, this is your reminder to get some water. Your PC can wait five minutes.
12
1000 Questions at once!
introducing the Gemini Decisions API
powered by Gemini Decision 1, our new multimodal decision model
get fast, calibrated decisions on up to 1000 questions at once across text, images, audio, video, and PDFs
try it today in AI Studio and via the Gemini API ai.studio
6
What if this becomes a natural part of Android messages...
Group chats will never be the same!
Introducing Tabletop, a collaborative space from @GoogleLabs where you can create custom tools to build, plan, and do more with your friends, family, or club.
Send an invite to your group and start a shared hub where everyone can contribute simultaneously for your next trip, potluck, or fantasy league. You can create live dinner polls, trip itineraries, expense splitters, and sports brackets tailored to your group.
3
Multi plan AI users,
Use Muse on the paid plan as your orchistrator for massive productivity boot.
We brainstorm and then it hands off different projects to all my CLIs and keeps me informed and makes suggestions as things run.
Have you tried it as your orchistrator yet?
10
Just found the best way to use Muse!
Just had 6 projects I was struggling with massively improved
The paid plan with Wonder is so much faster and amazing.
Tell your Muse you are going to brainstorm ideas and we need to flesh them out with options. And while doing that run audit subagents on every open project and give suggestions on how to improve.
Everything you like tell it to start a new chat for it and summon subagent swarm (thanks Antigravity) to build it. This keeps token costs down.
I burned 30 million tokens in an hour but wow I completed what took me at least 30 hours to start in 90 minutes!
42
Been saying Google is focused on cost effectiveness since Flash-lites came out. Now with a big model, looks like the same thing.
What are you going to use Carbon for?
What a day! I tested Gemini 4 Carbon on real work. 2 repos, 105 replanted bugs. Find and fix what you can.
It scores just behind Fable 5.2. The results:
- Fable 5.2 (max): 58 for $139.65
- Gemini 4 Carbon (high): 57.3 for $31.32
- Sonnet 5.5 (max): 51.3 for $154.03
- GPT-6 Astra (max): 45 for $33.03
- GPT-6.1 Sol (max): 44.3 for $5.88
- Gemini 4 Argon (high): 40.7 for $7.16
Google's own benchmarks looked super strong, but nobody could access the model. Turns out they didn't lie: +16.6 points over Argon, at less than a quarter of Fable 5.2's cost.
Nobody should be laughing at Google anymore.
(the next effort level drops soon in the thread 🧵)
14
Muse Ember score 100% on the benchmark to run in my budget!
P.S. Yes I know AI made up a few models. :0)
🤖 Made with AI
1
1
34
AI devs on a budget, you need to check out Muse Ember!
From HTML, WebGPU, low poly 3d, NPC logistics and object navigation, to sound effects. It has not let me down.
BREAKING: Muse Ember by @AIatMeta has redefined the Preference vs Speed Pareto frontier on Overall HTML, knocking out 10 top models from @GoogleDeepMind, @OpenAI, and @DeepSeek_ai.
At 33s per generation, Muse Ember is 9.5× faster on average than the previous Pareto frontier models. It is also at least 34× cheaper on average than all the previous Pareto frontier models it replaced at $0.10/$0.50 per 1M input/output tokens ($0.004 cache). It is also ranked 10th overall with an Elo of 1345!
A huge congrats to the @AIatMeta team on a top-10 placement on the leaderboard!
40
Muse Ember is going to change game dev!
Amazing low poly work,
Perpetual astroids,
Sound Effects,
NPC AI and navigation
That's the more impressive part! it wrote the script for the ship to auto navigate the field and miss them.
Muse Ember - high ~ $0.04
Has the basics of the game for less than $0.04!
Our first version of Muse Ember 🔥
Ember aims to maximize intelligence per GW of inference to help bring personal superintelligence to everyone.
We would love all the feedback to keep hill-climbing. Please send it our way. We will keep hill-climbing on efficiency, user experience, and better approach to train with internal frontier!
More to come 🫡
If you have not tried out Muse Code lately, it is an outstanding tool and with this new model the $5 plan will go a massively long way!
Seriously, give it a try.
Muse Ember is live 🔥. We built this smaller model for great performance at a fraction of the price. 35.8 AA Index at $0.14/task — ~3x cheaper than Haiku 5.5 at near parity. More intelligence per dollar than any of the models in its class.
Live on the Meta Model API and Muse Code, and with all our partners: OpenCode, OpenRouter, Cline, You.com, Cloudflare and more.
25
As a Parent AI Builder, it is hard to share what I do.
Only time I have is when I walk.
Used @GoogleAIStudio to make me a walk and talk screen recorder.
Would you use it?
Replying to @GoogleAIStudio
A Custom screen recording app that can detect a face and show the camera, when the face is not there it just keeps full screen. designed for walk and recording.
13
The biggest problem with @antigravity resets is my kids steal my computer and Antigravity again1
RL AI builder parent problems, for realz!
Resetting everyone’s Gemini and Argon quota on Antigravity.
Lots of updates this week so want to make sure everyone can build for the weekend. Lots more to come next week!
29




















