Fun setup to work on - and support the Phonon TTS pipeline via GenieX targetting the NPU it was fun to work on this in collab with @carrycooldude @ashwinexe @BhosalePratim

I got to build with @GradiumAI Phonon on launch day! A smart mirror that sees your outfit and gives spoken style recommendations - all running locally on @arduino Ventuno Q. @Alibaba_Qwen VLM to understand what you’re wearing Phonon TTS to talk through suggestions Both accelerated by the @Qualcomm_Dev Hexagon NPU

So cool to see @MLHacks giving out some @arduino UnoQs as part of prizes for their AI hackathon at @AllThingsOpen 🙌

Glad to be back at @AllThingsOpen after 2022! I am going to present a talk on how to improve security for running agents with @Docker Sandboxes - whether it is a coding agent or a multi agent system

An AI coding agent becomes useful when it can investigate a problem, change the code, test the fix, and follow through to a working deployment. That’s what I wanted to explore with the newly launched @reflection_ai Beam 🚀 Beam is built for coding, reasoning and agentic workloads. KubeTrace AI puts those capabilities to work on @kubernetesio incidents, with inference through @nebiustf. The agent can: → Investigate logs, events and application code → Trace a failure to its cause → Propose code or Kubernetes manifest changes → Run regression tests in a sandbox → Build and deploy the repaired application → Check whether it recovered The execution layer checks every proposed change against policy, and hidden checks decide whether the incident is resolved. You can inspect the evidence, patches, test output and failed attempts.

Featuring the Codex Ambassadors in India

Kicking off @OpenAIDevs DevDay Bengaluru

The problem with “YOLO mode” is that an agent can act with your access. A worktree won’t isolate your SSH keys, cloud credentials, or the rest of your disk. Jozu Agent Guard puts the agent in a microVM, applies tool-call policies, limits network access, and keeps an audit trail. I made this video with Opus 5.5 to show how the guardrails work 👇

Power of Local AI 😍 Running a custom Smart Mirror Demo application with @arduino Ventuno Q powered by a @Alibaba_Qwen 2.5 VL 7B model to describe what you wear, @googlegemma EmbeddingGemma 2 model for multi-modal vector search to find similar dresses, PiperTTS all targeting the @Qualcomm_Dev NPU on the Ventuno Q at IMC 2026. Stand in front of the mirror - and get recommendations on how to improve your outfit and get recommendations on other clothing options you can try all offline.

Shot a quick video covering @PrimeIntellect Agent and running it with the newly launched @Alibaba_Qwen 3.8-27B model on @nebiustf using nconnect. A few interesting features about Prime Agent: • A persistent Python environment for executing tools and managing working context using RLMs • A continual harness that can refine its supporting instructions based on experience The video covers the setup, how Prime Agent works, and the coding demo with Qwen 3.8-27B.

Reached highest loyalty tier of @TurkishAirlines Miles&Smiles program! Means free business upgrades for long haul flights to North America 😁

You ask a coding agent to fix a bug. A few tool calls later, it’s installing a dependency you’ve never heard of. Was the package name hallucinated? Coding agents make supply-chain decisions too. Here’s how @Jozu_AI Agent Guard can help 🧵

So impressed with the Opus 5.5 explainer videos - Here's another one - Serve Models from a @Kit_Ops ModelKit on @HAMiProject Detailed explanation of how Your model is packaged as a versioned ModelKit. KitOps pulls it from an OCI registry and unpacks it inside the Pod. HAMi schedules a controlled GPU share. SGLang loads the model locally and serves an OpenAI-compatible API. One clear path from model artifact to working inference. KitOps and HAMi.

Here's an video introduction to @tan_stack AI using a single prompt to generate a launch style video about Tanstack UI using @ClaudeDevs Opus 5.5 using @Remotion Prompt shared below 👇

I built WearScout because I kept seeing great outfits and spending way too long searching for similar clothes online 👕 Give it a reference image, it uses @deepseek_ai v4.1 Flash via @nebiustf to describe the look, and Jev drives a real browser using @browser_use to find similar pieces across stores, check product pages, and return a shortlist with prices and links. I tried it with a Tom Cruise outfit. I made a wrong call on privacy with my first demo, so I took that post down and changed the example. Heavily inspired by and adapted from @_nancychauhans’ Hearth project 🙌

this is great, you can run it with @nebiustf TF Relay now - nebius-tf-relay.vercel.app/

I built a very similar 3d world like GTA Style Gameplay UI andcompared @xai @grok 4.7 against Fable 5.1, Kimi K3, and GLM 5.3 Flash and pretty surprising to see how cost efficient Grok 4.7 is. While the UI levels / 3d artifacts don't match levels of Fable 5.1, it is significantly cheaper to run.

Got hands on with @arduino Ventuno Q - time for some on device VLM model inference 🎉 with @Qualcomm_Dev Hexagon NPU inference

I got early access to the new @Cline Open Source Desktop app! I wanted to see how it would fit into my everyday development workflow, so I tested it with the free GLM-5.3 Flash model across coding, scheduled reviews, and conversation handoffs. For the coding task, I asked Cline to build a small Node.js task-list CLI with persistence and tests. After resolving an initial environment issue and fixing how the CLI handled corrupted JSON, it successfully passed all 15 tests. But scheduling was the feature that stood out most. I scheduled a background review of the code and documentation. Cline compared the README with the actual implementation and caught a genuine data-loss risk: a corrupted file could be interpreted as an empty task list and later overwritten. That is the kind of work I don’t want to repeatedly remember to do myself. Overall, the foundation feels promising. The interface makes the agent’s plan and execution visible, scheduled tasks can remove repetitive work from the development loop, and importing work from another coding agent opens up some interesting workflows. Here’s my complete experience with the app 👇

Another guide with @rudrakshkarpe also in the works using @Kit_Ops to serve models with modelkits and serve them using @HAMiProject thus providing a safe way to securely get signed models with verifiable identity github.com/Project-HAMi/webs…

Just tried the latest @deepseek_ai V4-013 model using the DeepSeek harness through @nebiustf TF Relay. The entire setup and integration experience was seamless, and the model feels impressively quick in real-world use. It handled the task really well while maintaining a strong balance between speed, output quality, and pricing.

Spent 5+ hours debugging with GPT-6 Astra on medium, used ~40% of my usage, and still no confirmed root cause. Building an on-device Android agent with Gemma 4 E2B + Qualcomm QMX. live agent: ~2.6 tok/s. replaying the same request: ~11 tok/s. why the gap?😅

can agree - for me Goal Mode + Astra Medium has been effectively eating 1% usage every 5-7 minutes

What a wonderful start to @agentsnexus by @yashrajnayak and @sidagarwal04 Great discussions on multi agent architectures in production, security, monitoring and ofcourse on device AI agents @qualcomm_dev

I built an interactive 3D portfolio with GPT‑6 Astra: it turned my photo into a @Blender character, placed it inside an editorial WebGL experience, and created a revolving “Speaking Orbit” containing Earth globe mapping my talks around the world. The most impressive thing to note about Astra is how it moved quickly between web search, computer use for Blender and code. It created the 3D assets, integrated them into the website, and repeatedly inspected and refined its output. It also made this more refined video output.

Time to cook with @OpenAIDevs GPT 6 Astra 👨‍🍳

One problem I keep running into when building AI systems: which model should I actually use for each task? Benchmarks rarely give a complete answer. @withmartian AI Frontier looks at what matters in practice: - quality - reliability - real cost - performance across tasks The interesting takeaway: there isn’t one model that dominates everything. Increasingly, picking the “best model” is becoming a routing problem. Interesting data to explore: aifrontier.withmartian.com/

A wonderful start to the @OpenAIDevs Codex Bengaluru Meetup by @mvkaran sharing updates about Codex @blrtechweek @yashrajnayak

Ran the same GTA-style open-world prompt through three models: Kimi K3, GLM 5.3 Flash and @ClaudeDevs Fable 5.1. Rules: one HTML file, zero assets, everything procedural. Streets, traffic, pedestrians, day-night, enter a car and drive. Fable 5.1 nailed it in one go. No fixes, no retries, ran exactly as generated. Also the most polished city of the three by a distance. Kimi K3 needed a fix before it would run. GLM 5.3 Flash needed two, plus a repair pass to stop it crashing on load. The costs are the interesting part: Kimi K3 $16.89. Fable 5.1 $15.37. GLM 5.3 Flash $5.55.

Ran @Zai_org GLM 5.3 Flash and @Kimi_Moonshot K3 on the same task: build a GTA-style procedural open-world city in one HTML file. Zero assets, everything generated from code. GLM 5.3 Flash shipped more of the world, traffic lights, crosswalks, pedestrians on the sidewalk, seeded cities you can revisit with ?seed=N and did it for 3x less. $5.54 (GLM 5.3) vs $16.89 (KImi K3) for the same prompt. Both models shipped code that didn't run. Kimi's start screen couldn't be clicked, GLM had a null deref in its traffic AI. Neither caught it in 35 minutes of iterating, because the harness writes files and never opens a browser: neither model ever saw its own output. The difference was what happened next. I handed GLM the bug report and it repaired itself in 10 minutes for $2.09, then stopped on its own. Kimi never declared itself done at all. Ran both through @nebiustf using TF Relay.

Friendly reminder, even if @claudeai Models are down, you can still use Claude Code with OSS models like @Kimi_Moonshot K3 using @nebiustf In this video I show running CC with Kimi k3, with 1 step install using Nebius TF Relay. Your favorite model with your preferred harness

AI coding just got a lot more accessible for developers in India with @cursor_ai Start. They've launched a new India-only plan for ₹649/month (tax-inclusive), billed in rupees with UPI support. No more converting dollars in your head or wondering if an AI coding subscription is worth it. The video below shows Grok 4.5 building a SaaS product from scratch. It's seriously impressive for real-world coding, and you also get access to Cursor's other models, including Composer. If you've been waiting for an easier way to get started with Cursor, this is probably the best time.

It's available for the fine-tuning endpoint. Will be nice to have it as a public end point as well @demian_ai

A lot of us spend time obsessing over which model to use. I think the next optimization frontier is how that model gets executed. If your app can ask for "GPT-5.6-level behavior" (or another reference model) and let the inference layer figure out the most efficient execution while preserving the capability and behavior your application depends on, that's a much more interesting abstraction than hardcoding a specific model everywhere. That's why Ship caught my attention. Instead of building and maintaining your own routing logic, the endpoint optimizes execution at inference time while targeting the same quality grade, and does it at a fixed 50% lower price. The real win isn't just a lower bill; it's reinvesting that budget into more agents, longer reasoning loops, or additional verification.

Spain won the @FIFAWorldCup but this feels like another victory for @FCBarcelona. From Lionel Messi to Lamine Yamal, FCB has a legacy for producing class players. And when the biggest moment arrived, it was another Barça forward, Ferran Torres scoring

P.S. This is my first ever official reel. 😄 I procrastinated way too long on making a reel about my experience at Google I/O this year. Part of it was that I was waiting for the new release of @reelful by @katedeyneka Finally gave it a try, and I have to say, it's incredibly seamless to use. The way it understands the context across multiple uploaded clips and stitches them into a coherent story is genuinely impressive. For someone like me who's too lazy to edit videos, Reelful is pretty much the perfect solution. #GoogleIO

Really loving the exclusive #BuildWeek Swag from @OpenAIDevs Was such a great experience to host the @OpenAI Build Week hackathon in Delhi-NCR!

Since grok build is open source, I swapped @grok build backend for @nebiustf in one config file: ▪️ any OpenAI-compatible endpoint via [model.*] in ~/.grok/config.toml ▪️ @Kimi_Moonshot K2.7 Code as default, Qwen2.5-VL for vision ▪️ @tavilyai MCP for web search

I am subscribed to Maintainer Brief to get weekly updates about @mesheryio You can also get daily/weekly updates around your open source projects you maintain using the tool 🫡

Open source maintainer burnout isn't an effort problem. It's an attention problem. The signal is buried under hundreds of issues, stale PRs, and scattered HN threads. So I built Maintainer Brief, a weekly intelligence email that does the triage for you. powered by @nebiustf 🧵

cool app!!!!

Blame the new models for this cost in 4 days 😂

Continuing from yesterday's @cursor_ai benchmark here's where gpt-5.6-sol-high lands on tokens per dollar? ~1,971 output tok/$ ---> perfectly in the middle. It beats gpt-5.5 and opus-4.8-thinking-high-fast, but grok-4.5 and opus-thinking-high (~2,850–2,890) edge it out, and composer-2.5-fast (~6,950) is still untouchable on value. Performance wise, pretty similar to Fable for harder and long agent running tasks. P.S. personal run

Two Harnoors liking my post @iHarnoorSingh @Harnoor29581982 at the same time lol

I have tried out grok 4.5 quite extensively on @cursor_ai in a given repository where I also tested Fable, Opus 4.8 thinking high, 4.8 thinking-high-fast, GPT 5.5 and Composer 2.5 Fast. It was surprising to see that after multiple test runs,looking at the output tokens per dollar : claude-opus-4-8-thinking-high: ~3,245 tok/$ (44,426 output tokens for $13.69) grok-4.5-fast-high: ~2,955 tok/$ (15,427 output tokens for $5.22) So Opus wins by about 10% on output efficiency. While performance is quite similar in both. Not quite sure, if the benchmarks are slightly different, but for my usecase both Opus 4.8 and Grok are almost similar in both performance and pricing. Grok is not really an improvement in terms of cost. Takeaways: composer-2.5-fast is the clear winner — most output tokens for the lowest cost (~7,182 tokens/$). claude-opus-4-8-thinking-high is the best value among the Claude/high-reasoning tier (~3,902 tokens/$) and produced high output volume. grok-4.5-fast-high and gpt-5.5-high sit in the middle. claude-fable-5-thinking-high and claude-opus-4-8-thinking-high-fast are the least efficient (~605 and ~752 tokens/$) — highest cost paired with the lowest output. P.S. this is just a personal run, and output might be different for others.

Excited to share that @krisvelkov and I will be leading a workshop on Docker Sandboxes at @WeAreDevs World Congress 2026 in Berlin! 🇩🇪 With 13,000+ developers, it's one of Europe's biggest developer conferences. Sadly I won't be there in person, but looking forward to sharing the workshop with everyone!

At this point, we need a universal open source LLM router. One OpenAI-compatible endpoint + standard data scheme that aggregates all providers. Instead of every open-source project merging 50+ different provider PRs (most of which never get accepted anyway), they can just point to this central router. Who’s building this? 👀

And just like that, one of the greatest FIFA World Cup chapter of a generation comes to a close. For 20 years, Cristiano Ronaldo captivated the world with his brilliance. Thank you for everything CR7 ♥️🇵🇹⚽️

Building an ExecuTorch + local LLM Android app from a Mac can be extremely cumbersome. You end up fighting Linux-only build steps running inside @Docker , AAR/JNI runtime dependencies, ADB deploys, native crashes, device logs, and endless rebuild → install → test cycles. This is exactly where @OpenAIDevs Codex Desktop App shines for me. I connected it directly to my Android device over ADB while the app was being built from a Linux Docker container on my Mac, let it inspect real device logs and screenshots, defined the desired end state using /goal feature, and it kept iterating through the tedious debugging loop instead of me manually rebuilding, reinstalling, and validating every few minutes. ~5 hours later (and almost 50% through my session limit), I had ExecuTorch + @Qualcomm_Dev QNN running on the device.

This was super exciting to work on! Local first AI Agents on the edge are here! Got @goose_oss running completely offline on an Android phone. No cloud. No laptop. The Android app starts a local llama.cpp runtime itself, loads a @googlegemma 4 E2B GGUF model from device storage, and talks to it over 127.0.0.1. Flow: Android Goose UI -> app-managed llama-server -> llama.cpp / ggml -> @Qualcomm_Dev Hexagon HTP0 NPU target Local agent actions + local LLM + on-device NPU path. Tiny glimpse of where phone-native agents are headed. @techgirl1908 @AgenticAIFdn

A lot of @aiDotEngineer World Fair FOMO has been floating around my timeline today 😄 While I'm missing out this time, I'm grateful that just a few weeks ago I had the chance to speak at AI Engineer Melbourne about quantization and hardware-aware optimization for AI at the edge. Really appreciate that AI Engineer brings this experience to multiple cities around the world.

Hi @mervenoyann can we also add option for NPU?

And I thought my name was relatively unique...... 🫠 Someone beat me to it 😭 @WhatsApp

The pace of open-weight coding models over the past few months has been hard to ignore. What's becoming clear is that the model is only part of the equation. How the agent gathers context, uses tools, edits code, and recovers from failures has a huge impact on real-world coding performance. That's why the harness matters. ClinePass by @cline leans into that idea by bundling a curated set of open-weight models like @Zai_org GLM 5.2 behind a single subscription, so you're not constantly juggling providers, API keys, and rate limits. Couple this with a strong built in harness provided by Cline, and that is a win-win situations for devs. I recorded a quick walkthrough below

I didn't even realize my Maintainer Brief project was still quietly running in the background on my laptop. Today I got the daily email digest it generates every 24 hours. Seeing it land in my inbox was a fun reminder that it's been doing its job all along (showing @mesheryio)

As a maintainer, the hard part isn't code, it's knowing what matters. Which feature is truly gaining traction? Did a CVE just hit your deps? So I built Maintainer Brief. Every week it reads @github , Reddit, HN, and RFCs, advisories and emails you what changed: • the features actually gaining momentum • project mentions in reports • competitor analysis Built with @nebiustf, @resend and @nextjs

I’d been waiting for LiteRT-LM to support @Qualcomm_Dev X Elite NPU for @googlegemma 4, but that path still isn’t quite there yet. Turns out you can still target the NPU today using llama.cpp + Gemma 4 GGUF. Put together a quick Windows demo with llama-server, HTP0, and Gemma 4 E2B running locally. And it performs pretty well.

Same 😃

Had a wonderful time presenting a talk with @chamodshehanka on Flaky builds and self healing CI/CD using @opentelemetry at @KubeCon_ @CloudNativeFdn

Seems like @GeminiApp Code Review is pretty outdated. Gives a review that Qwen 3 doesn't exist 😅

If you are attending @KubeCon_ India, do drop by the Project Hami booth to see some great demos. An end-to-end GPU-sharing demo running on @kubernetesio : a single @NVIDIAAI A10 sliced by HAMi into a fractional vGPU (capped to 60%, the container sees 13.8GB of a 23GB card), serving @Alibaba_Qwen Qwen3-4B via @vllm_project. It ships as a @Kit_Ops Runtime Inference Container from @Jozu_AI Jozu Hub, one signed OCI artifact = server + weights baked in. So it's: kubectl apply → HAMi schedules it onto a sliced GPU → live OpenAI-compatible inference streaming back. Secure Model packaging + GPU virtualization, fully reproducible. 🧵👇@CloudNativeFdn

So excited to see @SaiyamPathak keynoting live demos at @KubeCon_ with Project Hami

Excited to be at @CloudNativeFdn @KubeCon_ in Mumbai! Here's where you can find me during the conference: 🎤 June 19 | 3:10 PM – 3:40 PM Talking with @chamodshehanka about Self-Healing CI/CD Deployments 🏛️ Project Booth Experience 🔹 Project Hami — Thursday, 3:30 PM – 7:30 PM (T-10) showing some demos using @Kit_Ops 🔹 Kserve — Friday, 2:05 PM – 4:10 PM 🔹 @kubeflow — Friday, 2:05 PM – 4:10 PM (T-7) 🔹 @mesheryio — Friday, 11:30 AM – 4:10 PM (T-13) Whether you're interested in AI infrastructure, MLOps, Kubernetes, platform engineering, open source, or just want to chat about developer experience, come say hi! And yes, continuing the legendary conference poster tradition by @TheTechMaharaj

piped.video/watch?v=Kd8uJx-O… Gave a talk on Quantum ML with @rishit_dagli back in 2023

Back to pre - LLM era, when the only conversations with Linus were about the Linux Kernel, and Git @linuxfoundation #ossindia Pic - OSS EU 2019 to OSS In 2026

At MCP Dev Summit Mumbai by @AgenticAIFdn and @linuxfoundation, @Anushka__Saxena and I presented a session on Rethinking Agent–Database Access using MCP Toolbox. A lot of agents get database access the dangerous way: a connection string and execute_sql. One prompt injection and your agent is exfiltrating every customer's orders. So we showcased a production-grade e-commerce agent that can't. The key is MCP Toolbox for Databases: @GoogleOSS open-source layer that sits between the agent and your database as a security control plane. The agent never sees a credential. Never writes raw SQL. It only calls narrow, typed tools Tech Stack: @PostgreSQL @MongoDB for Database, @Kimi_Moonshot K2.6 from @nebiustf for LLM Agent, Keycloak for OIDC, Runs locally on @Docker , deploys to GKE with Workload Identity Demo 👇

And it is pretty honest

I think it depends on the location you are in. Atleast in New Delhi, it is supposed to add the tolls. And that is how I always travel :)

See you folks there! piped.video/watch?v=1RwkiEuI… @kubesimplify @SaiyamPathak

Excited to be renewed as a @Docker Captain for another year! As AI coding agents, coding harnesses, and MCP-powered workflows move into production, secure execution environments and trusted infrastructure are becoming essential. Technologies like Docker Sandbox, Docker Hardened Images will play a key role in enabling the next generation of agentic applications. Looking forward to continuing to learn, build, and contribute with the container community! 🚀

This is great! I made a video showing how you can build self healing agents with Monocle using traces generated from errors to ask the agent to fix issues deterministically - piped.video/watch?v=FquPk9jt…

Friendly reminder: non-code contributions matter just as much in open source. Whether you're improving documentation, triaging issues, answering questions, or helping newcomers get started, it all counts. The best open source orgs recognize and value these contributions like @mesheryio So keep those contributions rolling - eventually those can also help you become maintainers🫡

For years, software delivery was constrained by one thing: Writing code. Now that's changed. AI can generate features, refactors, tests, and even entire applications in minutes. But there's a new bottleneck emerging: Confidence. How do you know the code actually works? Most teams still rely on a familiar workflow: → Generate code with AI → Run CI/CD checks → Review the PR → Merge The issue is that passing tests doesn't necessarily mean the user experience works. Broken flows, missing edge cases, UI regressions, and unexpected interactions often don't show up until real users hit them in production. That's why I found @Test_Sprite interesting. Instead of only analyzing source code, it launches your application and actively interacts with it like an end user would. The platform coordinates multiple AI agents that explore the product, build test scenarios, execute them, and document exactly what happened. A few capabilities that stood out: • Autonomous application exploration • AI-generated test plans • Replayable execution videos • End-to-end API visibility • Automatic adaptation to UI changes • GitHub workflow integration We're entering a world where creating software is becoming increasingly automated. The real challenge is verifying that what was created actually behaves the way users expect. That's the gap TestSprite is tackling.

Good friends at Kimchi are running a contest till 15th June to build anything. Kimchi is completely open source! This is a great opportunity to use Open Source Models like @Kimi_Moonshot Kimi 2.6 and an intelligent harness and model orchestration layer to build applications.

Youtube comments are always the funniest - Maybe even Opus 10 before GTA 6 🤣

Just tested Reelful and I have to say, it’s one of the best apps I’ve used for quickly creating reels. The AI does a surprisingly great job at identifying the exact moments worth highlighting. Everything from the editing flow to the final output feels thoughtfully designed. @katedeyneka 😄

Continuing to build with @googlegemma and LiteRT-LM Web API. Extended @carrycooldude’s Nova IDE into a @electronjs full fledged agentic desktop app with: - Real local filesystem access, - WebGPU-powered local inference using @Qualcomm_Dev GPU, - A new agentic coding mode.

Quick demonstration of the newly launched LiteRT-LM Web API with Gemma 4 E2B model running inside an Agentic IDE completely in the browser using WebGPU and @Qualcomm_Dev Adreno GPU. This new API provides built in prompt template for LLMs and I personally find it to be more optimized than the MediaPipe GenAI API, especially for agentic workloads.

@warpdotdev Not sure what's happening but Warp has completely shut down my laptop twice because it has been consuming 133 GB of RAM while running @ClaudeDevs Claude Code

Typical San Francisco Weekend: - Hike during the morning - Hackathon during the afternoon - Catching up with goated engineering friends during the evening

Participating at the #GoogleIO @GoogleDeepMind & @cerebral_valley hackathon at the amazing @SHACK15sf location in San Francisco, building with @GoogleAIStudio @antigravity and @GeminiApp

or sprinkling a bit of AI magic :)

Great to catch up with @ykdojo at Google Cloud and Nvidia meetup in Mountain View

What an amazing experience at the @antigravity Coffee co! A fully automated experience powered with @FlutterDev , Antigrativity and @GeminiApp to print custom Latte Art images on your drink Got to relive my favorite all time travel experience of seeing Northern Lights in Tromso

Catching up with @aaditsh on Day 2 At Google I/O

Highlight of the day - catching up with @OfficialLoganK at I/O

@jocarrasqueira sharing updates for Gemini and @antigravity and the entire agentic platform for development.