Philipp Schmid

5.9K posts

Philipp Schmid banner
Philipp Schmid

Philipp Schmid

@_philschmid

Agents & Gemini API, MTS @GoogleDeepMind | prev: Tech Lead at @huggingface, AWS ML Hero 🤗 Sharing my own views and AI News 🧑🏻‍💻 https://t.co/7IosdlO6RA

Nürnberg Katılım Haziran 2019
1.2K Takip Edilen79.2K Takipçiler
Sabitlenmiş Tweet
Philipp Schmid
Philipp Schmid@_philschmid·
Excited to introduce the Gemini Interactions API, a unified interface for Gemini models and agents. Starting today with Gemini Deep Research Agent. - Unifies access to models and agents via a single RESTful endpoint. - Access Gemini Deep Research agent via API. - Optional server side context management. - background execution for long-running inference. - Remote Model Context Protocol (MCP) servers support. Spent the last six months building this with an amazing team! I couldn’t be more proud and excited. Models evolve into systems and we want to give you the best developer experience. Give it a try and share your feedback. We will listen and want to build this with you in public.
Philipp Schmid tweet media
English
29
43
450
55.6K
Philipp Schmid
Philipp Schmid@_philschmid·
Gemini 3.5 Flash for video understanding and other modalities is super underrated! 🖼️🔊🎥
Philipp Schmid tweet media
English
6
0
57
4.8K
Logan Kilpatrick
Logan Kilpatrick@OfficialLoganK·
PSA: the IDE in Antigravity 2.0 is alive and well, we just landed an update to the UI which makes it more clear (see top right). Sorry for the confusion on this (pls keep feedback coming), we also just reset everyone’s weekly limits. Enjoy the weekend : )
Logan Kilpatrick tweet media
English
264
83
1.8K
112.4K
GunnerBAC
GunnerBAC@djpg2000·
@_philschmid I have Codex, Claude Code, and Antigravity/Gemini working together on projects trilaterally. Honestly, Antigravity/Gemini was having problems keeping up, hallucinating and/or lying, following orders, etc.. Let me know if you want more info. Not in front of my PC right now..
English
2
0
4
482
Philipp Schmid
Philipp Schmid@_philschmid·
Here are some third-party evals I came across for 3.5 Flash this week. It's doing well across agents, coding, vision, finance. Try it. Share what worked and what didn't. We will fix it. Every failure case makes the next version better.
Philipp Schmid tweet mediaPhilipp Schmid tweet mediaPhilipp Schmid tweet mediaPhilipp Schmid tweet media
English
10
11
130
9.9K
David Soria Parra
The release candidate for MCP 2026-07-28 is out. The protocol is now stateless: no handshake, no session id, any request can hit any server instance. Plus extensions as first-class (MCP Apps, Tasks), auth hardening, and a proper deprecation policy so we don't have to do this again. blog.modelcontextprotocol.io/posts/2026-07-…
English
70
233
1.6K
669.3K
Philipp Schmid
Philipp Schmid@_philschmid·
What if you can build an Agent with it own computer in a single api call? At my @Google I/O talk, I showed how to use Gemini Managed Agents and the new Interactions API to give your AI a secure, hosted Linux sandbox to execute code and manage its own memory.
English
17
27
283
20.8K
Luma5
Luma5@luma5b·
@_philschmid @Google So many solutions looking for problems. Google should make a better model.
English
1
0
0
224
Philipp Schmid
Philipp Schmid@_philschmid·
> Does the sandbox persist across interactions, or a fresh container per call? Yes it gets it own id, you can share it between different agents even. > Persistence vs cold-start latency is the usual tradeoff for long-running agents. At the moment we are slower. But this is our highest priority on improving at the moment. Cold start should be around ~7s, but the first over requests takes a bit longer. Please try it. Happy to hear more concrete feedback and needs.
English
0
0
5
380
Ömer Baş
Ömer Baş@omerbasdev·
@_philschmid @Google Does the sandbox persist across interactions, or a fresh container per call? Persistence vs cold-start latency is the usual tradeoff for long-running agents. Where does Managed Agents land vs E2B/Modal?
English
1
0
0
404
Regi Kusumaatmadja
Regi Kusumaatmadja@abdurrahmanregi·
@OfficialLoganK @_philschmid I don't have proof but gemini 3.1 pro extended in the app/web is definitely dumber than when I call directly via API. This mostly for math. No way systematic, but I can give you the same prompt and different in quality between the two if you want.
English
1
0
0
142
Philipp Schmid
Philipp Schmid@_philschmid·
Built a @github Issue Triage Agent with a single curl to the Gemini API. → Clones the repo into a sandbox → Fetches open issues from the GitHub API → Classifies each as Bug/Feature/Question → Executes reproducer code to confirm bugs. No orchestration framework. No containers. No infra.
Philipp Schmid tweet media
English
8
6
63
5K
🚨 AI News | TestingCatalog
ICYMI 👀: Google AI Studio got a new "antigravity-preview-05-2026" Agent! > A general-purpose autonomous agent running in a remote, Google-hosted Linux environment. > This agent can execute code, take real actions, and use a large number of tokens.
🚨 AI News | TestingCatalog tweet media
English
13
14
305
16.7K
Khushboo Verma
Khushboo Verma@khushbooverma·
I sat down with @OfficialLoganK for an honest technical breakdown of the Google I/O updates. 🎙️ We pushed on: • Gemini 3.5 Flash pricing • Gemini CLI deprecation • Benchmark gaps vs Opus & GPT-5.5 Huge thanks to Logan and the DeepMind team for leaning into the tough community feedback and giving us real engineering clarity.
English
42
25
498
76.3K
Simon Willison
Simon Willison@simonw·
Anyone understand what Google mean by "Gemini Spark runs on Gemini 3.5 and uses the Antigravity harness" - is "Antigravity" a generic term they're using for their agent harnesses now or is their Claw-competitor running the same closed-source Go binary we can download ourselves?
English
58
7
223
31.2K