Harshit Rathi

iX1 sales more than X1! That's commendable conversion from the brand towards EV. #bmw

You can now ask Gemini to create Docs, Sheets, Slides, PDFs, and more directly in your chat. No more copying, pasting, or reformatting, just prompt and download. Available globally for all @GeminiApp users.

India doesn't need to lead the world in building the most advanced AI models. But it must lead in ensuring benefits of AI are widely shared. @rvenk and I have an op-ed in The @EconomicTimes economictimes.indiatimes.com/opinion/et-com…

Express Adda- Ruchir Sharma: my quick takeaways - US leads the AI race, powering its markets and economy - The world criticizes America, but capital continues to flow there - TSMC (Taiwan) thriving; Samsung will earn this year more than any US company except Nvidia - Americans largely insulated from war impact; Asia feels it more - China has cleverly built strong energy reserves - The world is burdened with excessive debt and deficits - India faces headwinds: limited AI scale, war spillovers, possible El Niño impact- yet India’s resilience remains a strength. Bottom line: India must aggressively attract FDI, build AI capability, and secure energy to stay relevant in the new global order.

@avgspacelover Ya, the new multipliers are insane.

@HarshitRathi_ Its 27 for 4.7 now

Github copilot needs to sort this out... 7.5x limits for frontier models like opus4.7 and gpt-5.5 is too much.

Agent Mode is here in Outlook! Copilot can now help run your inbox and calendar, triaging emails, rescheduling meetings, and helping you stay on top of what matters most.

WHAT HAVE WE JUST WITNESSED? 🤯 Sabastian Sawe has just become the first person in history to run a sub two-hour marathon in race conditions. Yomif Kejelcha was also under two hours for second!

Codex with gpt 5.5 is super good at instruction following and Agentic coding.

feels like a good time to seriously rethink how operating systems and user interfaces are designed (also the internet; there should be a protocol that is equally usable by people and agents)

India's peak power demand hit 260 GW yesterday. New all-time record. During a heatwave with 47°C temperatures across North India. The grid held. No blackouts. That sentence alone is an engineering achievement most people will not appreciate. Five years ago India had chronic power deficits. Load shedding was normal. In 2023, peak demand hit 243 GW and several states struggled. In April 2026, the grid delivered 260 GW without a single major failure. What changed: 26.5 GW of new capacity added in FY26 — largest annual addition in a decade. Solar alone contributed 18 GW. New HVDC transmission corridors connecting surplus regions to deficit ones. Battery storage deployments cushioning peak load. But the margin is razor thin. India's grid is designed for about 270 GW. We just touched 260 GW. That is 96% utilisation during peak hours. One more heatwave spike or an unexpected plant outage and the buffer disappears. This is why every power stock hit 52-week highs. The market sees what the headlines miss — India needs $150 billion in power infrastructure investment over five years just to keep up. Data centres, EVs, semiconductor fabs, industrial expansion — all need reliable 24/7 power. The grid is the bottleneck holding everything else together.

Mac minis with 32GB+ have a 10-18 week wait right now. PMs are buying them as personal AI compute boxes. A developer in Australia named Peter built something called OpenClaw on a weekend. The pattern is dead simple. You install an agent on a dedicated Mac mini sitting in your closet. You message it on WhatsApp. It runs the work on the Mac mini. It sends the result back through WhatsApp. You never open a terminal. You never sit and watch it think. The agent has full control of one machine that isn't yours, with full bash access and full file system access, sandboxed away from anything that matters. Three things make this different from Claude Code: Delegation through channels you already use. WhatsApp, Slack, email, SMS. The agent works while you sleep, eat, or run errands, and pings you when it's done. You stop being the bottleneck. Full machine sandboxing. Instead of granting file-by-file permissions every time, you give the agent its own computer. The Mac mini is the sandbox. If the agent does something destructive, it destroys a $599 machine, not your work environment. Model agnosticism. Connect any model, including open source. No Anthropic rate limits, which is the single biggest complaint from heavy Claude Code users. The reason this matters at enterprise scale: the same architecture is what GCP and AWS are about to ship inside their cloud platforms. Send a message to a sandboxed agent that reproduces your problem, tries a solution, returns the result. The Mac mini is the early indie version of what Google is going to sell as a managed service in 2026. Mahesh Yadav has been running this setup for months. 13 years building AI at Microsoft, Amazon, Meta, and Google before going independent. His take: PMs who learn the OpenClaw pattern now will recognize the shape of every enterprise agent platform when they arrive. The hardware shortage is a leading indicator. Builders are buying compute before companies are.

Instead of watching a movie, Learn how Andrej Karpathy (@karpathy), OpenAI co-founder, ex-Tesla AI, "vibe coding" creator explains why Claude Skills, MCP servers, and AI agents are past the hype and are now the new baseline for building in just 262 minutes ↓

Claude is miles ahead of any other AI app right now. The number of internal tools our team has built over the past few months is honestly insane, both in depth and detail.

Apoorva is one of the best founders I’ve ever met in my life. This is a big big idea.

Twitter acquisition price: $44B Cursor acquisition price: $60B 🤔

We need MacBook with Opus-level model running on it locally, for free, without limits.

This 30-min workshop by the creator of Claude Code will teach you more about vibe-coding than 100 YouTube video guides. Bookmark it & give it 30 minutes today. This video will change the way you use Claude forever.

Did xAI just mass-murder the entire voice AI industry? 🤯 Grok just launched two voice APIs. Speech-to-Text and Text-to-Speech. Built on the same stack powering Tesla cars and Starlink support. And priced at 10x cheaper than ElevenLabs. Speech-to-Text: $0.10/hr batch. $0.20/hr streaming. Text-to-Speech: $4.20 per million characters. 25+ languages. Real-time streaming. Speaker diarization. Already outperforming ElevenLabs, Deepgram, and AssemblyAI on word error rate. TTS ships with expressive tags like [laugh], [sigh], , . Voices that don't sound like robots reading a script. ElevenLabs spent years building a voice AI company. xAI built voice AI for cars and satellites.

Here is what I tell our software engineers on how to thrive in the AI era: be very good domain experts. Programming skills are the foundation (and we definitely don't want to lose them) but deep domain knowledge is what customers pay for, along with reliability, security, support and compliance. The productivity gains from AI are still hotly debated: we definitely get to a working prototype much faster but a finished product has a lot more to it and not all the stages can be sped up by AI. That is why I advise our technical teams to not obsess about programmer productivity as a metric but focus on how we can offer a far better experience to customers using AI. There is a lot of needless or incidental complexity in software that can be eliminated by AI.