Simba (Egide)

$6.3B stolen in about 20 minutes Binance got the doors back open in 3 days The hacker spent this Sunday morning turning $95.9M of SOL into BTC, live and fully traceable The blockchain is a perfect crime scene camera Nobody can arrest a wallet though

What the people have been waiting for Grok @bot + @cursor_ai + @X is an unbelievable stack to build anything your heart desires

Grok @bot is honestly an insane product

we're going to see a whole new generation of builders who've never seen or written a line of code build incredible things

Grok @bot might be the most important AI release of 2026 A few days ago it felt like a niche toy. Now it routes to Opus 5.5 for harder work, runs much faster, and has full access to X People are already using it to triage mentions, ship features, and handle real ops. You can build an entire company with it…support, research, outreach, content, and engineering, all running as agents that keep working while your laptop is closed Bravo @spacexai @poteto @elonmusk

Holy, this is already live. @bot now routes every task to the best model for the job, xAI's or third-party Announced yesterday, shipped today. Wild pace

If the cost of software engineering is going to zero, the only moats left are creativity and speed to market

Everyone keeps asking why crypto still feels hard to use Maybe it’s because we keep trying to make humans the user Wallets. Seed phrases. Gas. Signing every little thing. That UX isn’t “early.” It’s what you get when the rail was built for machines that never forget a key, never sleep, and settle 10,000 times a day without opening an app Humans needed banks, receipts, and customer support Machines needed a shared ledger that doesn’t ask who you are…just whether the math clears The weird part: once agents start paying each other for compute, data, and API calls… the “confusing” interface suddenly looks native We didn’t fail to make blockchain human. We just built it for the next customer.

An underrated feature of Grok + Cursor is the video demos for built features The bot doesn’t just open the PRs, it can hand back a short clip of the change actually working That makes the handoff much easier to review

I think we’re starting to measure AI the wrong way For the first few years of generative AI, most of the conversation was about output quality. Which model writes better? Which one codes better? Which one reasons better? Which one makes better images? That made sense when AI was mostly something you prompted and then reviewed Agents change the equation If software can research, make decisions, use tools, move through workflows, talk to other agents and eventually transact on your behalf, then the more important question becomes: how much actual economic work can it complete? And I think that leads to a metric we’ll hear a lot more about over the next few years: How much verified economic work can one human safely supervise? Today, one person might manage a few agents doing research, coding or operations. Over time, that could become dozens or hundreds of specialized digital workers operating in parallel At that point, headcount alone tells you much less about the productive capacity of a company. A 20-person company with the right systems could potentially execute like a company many times its size. The real AI leverage won’t come from generating more content It’ll come from turning human judgment into machine-scale execution

If you are not using grok bot what are you even doing bro?? 🤔

Agency isn’t motivation Motivation is a feeling Agency is a decision you keep making when the feeling leaves It’s shipping the ugly first version It’s sending the awkward message It’s pointing the agents at “your” problem instead of scrolling someone else’s build In a world where tools can execute for you, agency is choosing the target and not negotiating yourself out of it Godspeed on your endeavors

Everyone’s racing for better models I keep noticing a quieter race You can spin up agents (Grok, Codex, Claude), pull almost anything off the internet, and have workflows execute while you sleep Knowing got cheap Doing got cheap What’s left is whether you actually want something badly enough to point them at it Imagination + agency is the whole game again What a time to be a builder

GPT-6 Astra was just released I went through OpenAI's launch notes, API docs, safety research, benchmarks, and early customer tests Here are the parts I think actually matter: The biggest improvement isn't that AI got smarter It's that AI needs you less. Astra can research, use software, navigate websites, work across files, write code, create documents, test its own work, and carry multi-step jobs toward a finished result It can even keep working while waiting on another tool and let you change direction while the task is already running That sounds small. I think it's a massive shift. We went from: AI that answers questions → AI that helps you do work → AI you can increasingly give the work to The early examples are interesting. One company reported 50% fewer manual fixes. Another had a single agent review 41 financial documents in minutes and catch every planted error. But here's the part I'm watching: The UI is becoming an API. If AI can use software the way a human does, millions of workflows no longer need a perfect API or custom integration before they can be automated. And that has another consequence Your SOPs become instructions for agents Your templates become executable standards Your company knowledge becomes infrastructure The companies with the cleanest operations may suddenly become the easiest companies to automate So I wouldn't obsess over Astra's benchmark scores. I'd ask a different question: What workflow inside my business can I now hand AI from beginning to end? Because the real AI race is shifting from intelligence to execution And once AI can reliably own the workflow, it stops being software you use. It starts becoming labor you deploy

I’ve noticed more top founders moving deeper into AI infrastructure, and I think there’s a simple reason why The obvious opportunity in AI is to build applications. But applications are getting easier to create, features are being copied quickly, and models themselves are becoming more accessible. That shifts the question from “what AI app should I build?” to “what will every AI company eventually need?” Things like compute, data, payments, security, memory, agent infrastructure, and tools that help AI systems actually interact with the real world We’ve seen this before. During the internet boom, the flashy consumer products got most of the attention, but a lot of durable value ended up in the infrastructure underneath them AI could follow a similar path There may be thousands of AI apps over the next few years, but only a small number of infrastructure companies powering all of them That is the part I think more founders are starting to recognize

One of the more interesting things that happened this week Cisco giving all 90,000 employees their own AI agent But I think people are missing the bigger story. The basic unit of a company is starting to change For most of history, more productive capacity meant hiring more people. AI agents break that relationship One employee can increasingly manage research agents, sales agents, analysts, support agents and operational agents working underneath them So a 50-person AI-native company could eventually have the execution capacity of a 500-person traditional company That's why I don't think the future is simply "AI replaces employees." Employees themselves are becoming managers of digital labor And once that happens, headcount stops telling you how powerful an organization actually is

The real shift isn’t “ChatGPT can now log in.” It’s that we’re giving agents their own persistent computer and a secure identity that can act across the web without ever seeing your credentials. Trip planning used to mean 12 tabs, 4 logins, and you still doing the last 20%. Now the last 20% is also the agent’s job. This is the beginning of work that happens whether you’re looking at your phone or no