Ceebo

[Google @Google]

Today we’re introducing Gemini 4 Argon.

It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.

[Mr Ash @ash_twtz]

Just noticed something interesting with Antigravity:

It automatically generated an image with Nano Banana while building my website.

Imagine where this goes if Google nails the next step:

- Generate the images
- Generate the videos with Veo
- Automatically integrate everything into your site
- All natively, in one agentic workflow.

Even a $200 Claude Code Max plan can't do this natively. It can only orchestrate external tools, but doesn't have this native image + video generation stack.

If Google pairs this with a truly great coding model, Antigravity could become a serious contender for the best AI code editor.

[Colaboratory @GoogleColab]

Colab is now part of Google AI plans! Subscribers get priority access to faster accelerators and more powerful machines, Ultra users also get access to longer running background execution and Premium GPU access.
Learn more here:
https://t.co/zeUeqjVOVg

[Ceebo @KilaArik]

@rodydavis I like antigravity very much, but I really need a built-in pdf browser like codex.

[Ceebo @KilaArik]

@burkov You mentioned Gemini 3.7 Flash, I would like to know where the comparison of Gemini 3.8 Flash stands in your case?

[Ceebo @KilaArik]

Can not agree more.

[Uzer Kayat @UzerKayat]

A suggestion for the @antigravity team.

I’ve been using Antigravity and I keep thinking about one direction that could make the product much more interesting.

Personally, I’d rather see Google remove the third-party model options like Claude and GPT-OSS from Antigravity and use that space to bring more of Google’s own AI models and capabilities directly into the app.

Google already has a huge AI ecosystem. There’s Gemini for reasoning, coding and agents, the Nano Banana family for image generation and editing, Gemini Omni Flash for video generation and conversational video editing, Lyria for music generation, and other generative capabilities available across Google AI Studio.

Imagine having all of that properly integrated into Antigravity.

Of course, Google could put separate rate limits on image, video, music and other expensive generation models depending on the user’s Google AI plan. They wouldn’t necessarily need to make everything unlimited.

I understand why Claude and GPT-OSS inside Antigravity having is useful for people who want model choice, but I think there’s an even bigger opportunity here. Antigravity could become the place where the entire Google AI ecosystem comes together instead of being another app where you simply choose between different LLM providers.

One app for agents, coding, research, image generation, video creation, music and whatever Google build next.

Google already has most of the pieces. Bringing them together could turn Antigravity into a genuinely powerful Google AI super app.

Would love to see the @antigravity, @GoogleAIStudio and @GoogleDeepMind teams explore this direction.

[Ceebo @KilaArik]

Antigravity中的Gemini 3.8 在昨天经过高负载崩溃和官方修复后的额度重置之后，在今天又变得非常慢...

[Ceebo @KilaArik]

@rodydavis I often need to do paper writing work, so I hope Antigravity can open PDFs in the sidebar and display them normally. Currently, when I open it, the PDF displays as garbled text, which is very troublesome for me.

[Ceebo @KilaArik]

@Lonely__MH 慢死我了

[Google AI Studio @GoogleAIStudio]

https://t.co/5g6C516PN3

[Ceebo @KilaArik]

真的受不了了，Gemini今天一整天为什么变得这么卡了？大的要来了吗？

[chuplung @choopyplug1]

Anthropic just dropped a 13-page PDF on Agent Memory - 5 layers that cut token cost 90% and make your agent actually learn:

here's the 5-layer memory architecture:

layer 1 → working memory - the context window. everything the agent sees right now. when it fills up, old context dies. most agents stop here and wonder why they're broken

layer 2 → episodic memory - what happened. full interaction logs with timestamps. the agent recalls that the deploy failed Tuesday at 3am because the migration script had a typo

layer 3 → semantic memory - what is true. facts, entities, relationships stored as a knowledge graph. "user prefers TypeScript" lives here. doesn't expire when the session ends

layer 4 → procedural memory - how to do things. the agent tried 3 approaches, one worked. that method becomes a reusable skill. next time it skips straight to what worked

layer 5 → forgetting - what to delete. an agent that never forgets accumulates contradictions. old preferences override new ones. the user moved cities but the agent still recommends restaurants in the old one

the result: Mem0 stores 1,800 tokens per query instead of 26,000. Snowflake added one ontology layer - 20% better accuracy, 39% fewer tool calls. memory pays for itself on day one

this 13-page PDF is what separates a chatbot from an agent that actually learns

don't scroll past this one ↓

[Ceebo @KilaArik]

@kyuuurius @GeminiApp @antigravity me too

[John @johnAGI168]

Antigravity 真是越用越好用啊😄

AI 剪辑一条龙基本上全部搞定！在谷歌 AI Studio 复制 API 密钥，存到钥匙串，就可以在Antigravity 里面调用谷歌的所有模型，包括 Nano Banana 还有音乐模型Lyria 3.5。

这样从剪辑一条龙：从粗剪、精剪，最后封装，调用音乐模型生成合适的 BGM，再调用图片模型生成视频封面，完美！

[Ceebo @KilaArik]

@rodydavis 我真的非常需要侧边栏内置 pdf 阅读器的功能，拜托了！😢

[Ceebo @KilaArik]

@BystAnd3rs 相比DeepSeek harness怎么样？

[Ceebo @KilaArik]

Harness: PI vs Deepseek ?

[Ceebo @KilaArik]

@robotbird01 对不起，不算很了解。为什么看到好些都是选用 PI，而不是 Deepseek 的 harness？

[Ceebo @KilaArik]

@ego_agent REPORT

[Aymane @GamefxAI]

Want to make a game with AI?

Delete Unity.
Delete Unreal.
Delete Godot.

They were never designed for AI to work with. Use Three.js instead. LLMs suck at heavy engine UIs, but they excel at writing WebGL -> playable browser link in 30 seconds.

@threejs @webgl_webgpu