Hamid Shahid

"If you want to go fast, go alone. If you want to go far, go together." Let's connect and grow far together. 👋🏻

we're starting rollout of GPT-5.5-Cyber, a frontier cybersecurity model, to critical cyber defenders in the next few days. we will work with the entire ecosystem and the government to figure out trusted access for cyber; we want to rapidly help secure companies/infrastructure.

Check it out on GitHub: github.com/openai/realtim…

The future of "Hands-Free" UI is officially here. OpenAI developer Jason Liu just showcased the power of the gpt-realtime model, and it’s a game-changer for how we interact with web apps. No screen scraping—just pure, low-latency natural speech. Dynamic Control: Toggled dark mode and navigated UI via voice. Contextual Forms: Filled out complex forms by simply dictating details. Real-time Gaming: Played a full game of chess by voicing moves like "Knight to F3." While the community noted minor "jargon" hiccups and audio loops, the speed is undeniable. We’re moving away from clicking and dragging toward a world where apps simply listen and react. From kids' phonics apps to accessibility tools, the "Voice-First" era is gaining serious momentum. 🚀 #OpenAI #RealtimeAPI #GPT4o

@0xsiddhant_ @oprydai I am using Ubuntu. It has good UI What about you?

@ch_hamid77 @oprydai What distro are you using ?

do people still use Windows?

The real story in Google's up to $40B investment in Anthropic isn't just the money; it's the tightrope walk of being partners and rivals. $10B is immediate cash, with $30B contingent on hitting performance targets at a $350B valuation. This move, explicitly supporting a significant expansion of Anthropic’s computing capacity, deepens a strategic tension unprecedented in AI. What kind of internal firewall do you build when you're directly funding your top competitor's compute growth?

My first book is the biography book "My brother" by Fatima Jinnah. She wrote about the life and political struggle of her brother Muhammad Ali jinnah, the founder of Pakistan. The book was published after her death. Book is still a controversial that the material in this book may be tempered by government authorities.

The era of "prompting" is over. We’ve entered the age of agentic orchestration. With GPT-5.5, OpenAI is officially moving the goalposts. The transition from the Assistants API to the Responses API by 2026 isn't just a rename; it is a total infrastructure shift toward autonomous compute. Key takeaways for devs: Native Control: hosted shell and apply patch mean models are no longer just writing code; they are managing environments and surgically updating filesystems. The Agents SDK: A blueprint for building "Harnesses" and "Skills," making durable, long-running agent execution a production standard. Architectural Shift: Moving from "Runs" to "Conversations" allows for granular control over reasoning and tool-calling loops. The question for 2026 isn't what the model can say, but what permissions you’re ready to give it. Is your stack ready for a model that runs its own terminal? #AI #OpenAI #AgenticAI

Users are cancelling Claude Pro, not just for declining quality, but for systemic token issues and abysmal support. Users experienced 100% token spikes on simple questions, vanishing conversation caches forcing costly re-reads, and automated support that closed tickets without addressing the problem. Are you facing the same issue with Claude?

For engineering & product teams: If you're building at the cutting edge of AI, the TPU 8 series warrants evaluation for your next infrastructure cycle. What workload would you prioritize first: training scale or inference efficiency? #CloudComputing

Why it matters: → Million-chip clustering unlocks unprecedented scale for frontier model training → TPU 8i's memory architecture accelerates complex reasoning & multi-step agentic workflows → Significant efficiency gains reduce TCO for large-scale AI deployments

Google announces next-gen TPUs: - TPU 8t: Training frontier models | +124% perf/watt | ~3x compute/pod vs Ironwood | Scales to 1M chips - TPU 8i: Inference & agentic AI | +117% perf/watt | +80% perf/$ | 288GB HBM + 384MB SRAM/chip GA later this year. #AI #MachineLearning

@cursor_ai It is incredibly promising to see two world-class technologies, SpaceX's Colossus supercomputing and Cursor's AI coding expertise, working in concert to make AI-assisted development faster, smarter, and more accessible than ever.

We're partnering with SpaceX to improve Composer. cursor.com/blog/spacex-mo…

This partnership signals a new paradigm: combining Cursor's deep integration with professional engineering workflows and SpaceX's unprecedented AI training infrastructure to build the world's most capable coding AI.

The rapid growth of the Codex user base, highlights a fundamental shift in the software landscape. This trend reflects the increasing leverage of AI-powered development and 'vibe coding,' underscoring how deeply generative tools are becoming embedded within industrial workflows.

A significant transition for Apple as Tim Cook steps down as CEO. During his extraordinary tenure, Tim didn't just scale a company; he championed privacy, sustainability, and operational brilliance. Thank you, Tim! for steady, principled leadership, and look forward to the future under John Ternus' vision as the new CEO. #Apple #TimCook #JohnTernus

Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.

@oprydai Beautifully said!...math has always been one of my favorite subjects since I got into the computer Science field.

math is the gap between imagination and reality. - splashdown

Home sweet home 🌍

Anthropic's Project Glasswing is the most interesting AI launch in a while. They built a model so powerful at hacking that they refused to release it publicly. Instead, they quietly gave it to Amazon, Apple, Microsoft, Google & 40 other orgs with one job: find every critical vulnerability before the bad guys do. Results so far? → Thousands of zero-days found → A 27-year-old bug in OpenBSD → A 16-year-old flaw in FFmpeg that ran 5M times undetected → Chained Linux kernel vulnerabilities to gain superuser access This isn't a chatbot. This is AI doing the work of entire security teams automatically. The scary part? This is the defensive use. $100M in credits. $4M to open-source orgs. And a model they won't let the public touch. We're in a different era now #claude #AI #glasswing