Gemini 3.5 Flash gains native computer use

#1 𝕏

Google DeepMind has added native computer use to Gemini 3.5 Flash, giving developers a built-in tool to build custom agents with vision and action capabilities across browser, mobile, and desktop interfaces.

#2 𝕏

Mustafa Suleyman launched MAI-Image-2.5, now ranked #2 for text-to-image and #3 for image editing on @ArtificialAnlys (just behind GPT models), alongside MAI-Image-2.5-Flash for best-in-class quality/price.

#3 𝕏

OpenAI unveiled its Agents API, powered by GPT-4 Turbo’s function calling, memory, and tool integrations, enabling teams to build custom AI agents for tasks like document summarization, CRM updates, and meeting scheduling.

#4 𝕏

Philipp Schmid launched computer use in Gemini 3.5 Flash across browser, mobile, and desktop environments.

#5 𝕏

Harrison Chase outlines the creation of SmithDB, a database purpose-built for storing and querying agent traces, detailing its architecture, specialized features, and performance tuning.

#6 𝕏

Guillermo Rauch shows how they’ve built AI coding agents that automatically enforce Vercel’s design standards by integrating Figma tokens, Tailwind CSS classes, and component guidelines directly into the code-generation pipeline.

#7 ▶️

Connect 9,000 Apps to One Unified AI Dashboard (Claude Code + Zapier SDK)

Helena Liu

#8 ▶️

"Learn AI” Is Bad Advice. Learn This Instead

Greg Isenberg

#9 𝕏

Cursor rolled out constrained eval environments (sandboxed I/O, fixed test seeds, no network) to curb reward-hacking in coding benchmarks and ensure scores accurately reflect model intelligence.

#10 𝕏

There's An AI For That lays out a 3-step AI system design—decouple planning from building to prevent context rot, keep static rules minimal and load skills on demand, and fix system-level issues instead of patching model bugs—arguing that the surrounding system, not the model...

#11 𝕏

LlamaIndex 🦙 built LiteParse, the fastest open-source document parsing solution on the planet, and it just surpassed 10k stars on GitHub.

#12 𝕏

Ali Ghodsi shows that open-source GLM runs at over 300 tokens/sec versus ~100 tokens/sec for proprietary frontier models, delivering a 3× speedup that’s critical for responsive agentic workloads.

#13 𝕏

Google Research unveiled Linear Elastic Caching, framing page eviction as a ski rental problem and using lightweight ML to optimize the memory-footprint versus cache-miss trade-off, cutting total cache costs.

#14 𝕏

Santiago breaks down agent improvement into three areas—model (for code/math), harness (tools and safety checks), and context (plain-text logs)—but argues that nothing beats continuous learning from every user correction.

#15 𝕏

xAI integrated SuperGrok and X subscriptions into T3code, letting developers access their AI-powered features directly within the coding environment.

#16 𝕏

Logan Kilpatrick says Gemma 4 is best-in-class across all hardware tiers, designed not for server-side frontier AI like GLM but to enable local, on-device intelligence without needing advanced hardware.

#17 𝕏

Teresa Torres launched “Is This Okay?” (ITO), an AI-driven teen consent reflection tool from Override Labs designed to help prevent sexual assault before it happens—without tracking users, judging them, or issuing verdicts.

#18 𝕏

claire vo 🖤 invites you to watch her full talk on YouTube and thanks @leerob for the invitation.

#19 𝕏

Thariq recommends using workflows and adding a second agent to verify outputs and provide feedback for this use case.

#20 in

Peter Yang warns that strict identity verification is 100% coming for frontier AI model access, signaling a major shift in how these systems will be secured.

#21 in

Peter Yang offers a beginner-friendly, 45-minute live tutorial to install the Hermes app, configure a voice-enabled Telegram bot, connect Google Workspace, and create custom routines.

#22 𝕏

There's An AI For That shifts you from coder to “factory manager”: you draft the system, specs, and guardrails, and the AI agent then outputs the code, docs, and tests within your framework.

#23 𝕏

Anthropic joined @raiseus_ai as a founding partner, supporting the nonprofit coalition’s effort to strengthen the American workforce through employer-led initiatives, AI-powered training, and policy innovation for transformative AI.

#24 𝕏

Philipp Schmid publishes a hands-on guide to integrating Google Gemini in Android apps, detailing setup steps, API calls and example code.

#25 𝕏

OpenAI teams across every department are deploying Codex-powered agents to handle complex, long-running, cross-functional tasks. This internal rollout offers an early glimpse of how agentic tools may reshape work as they become more capable and widely available.