Alex Grankin

I spent years at McKinsey helping organizations make decisions. Then I started building with AI — and realized most smart professionals are stuck at "I've tried ChatGPT." I'm building @theaibridges to fix that. Practical AI education. No coding required. Just what actually works. Follow along.

@JeffKirdeikis Don’t even get me started with 1-1.5 hour bookings appearing in the calendar

Finally an agentic mobile browser from @perplexity_ai

@faheem_munshi @owner Gotcha, so you create AI slop, makes sense!

I send my team a Slack update every Friday. I've never typed a single one. Every Friday at 4pm, Claude Cowork automatically: Reads my week notes file Writes the full report (wins, KPIs, action items) Posts it to #team in Slack as @owner | task | due date Archives it to /Reports/ — named and dated The prompt is 9 lines. The /schedule command is 1 line. Setup took 10 minutes. That was 4 months ago. Hasn't missed a Friday since. Most people with Claude Pro have no idea this exists. #ClaudeAI #Automation #WorkSmarter #AIProductivity

Co Claude Cowork is finally taking on the OpenClaw 😁 trying it out today!

@felixrieseberg When will I be able to connect Claude to my mind. Guys…

We're shipping a new feature in Claude Cowork as a research preview that I'm excited about: Dispatch! One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work. To try it out, download Claude Desktop, then pair your phone.

Everyone is talking about how fast Claude is shipping. So I thought to myself - when was a significant ChatGPT update that I noted? I don't think there was one ChatGPT app update in a while. Or am I missing something?

@MarcusTull20019 So painfull... Just saw that Opus 4.6 is down. How are we supposed to live on Sonnet 4.6...

@grankin I feel elite now

You know you're a top adopter when you find out that Claude Code has systems down before status.claude.com shows it

@JeffKirdeikis Anyone can be a successful solo entertainer, when you have autonomous company working with you

Autonomous companies outperform human-powered on every metric that matters. Speed, Cost & Scale The future is clear, placing investment bets has never been as easy as it is today.

I stopped uploading files to AI apps. Not on purpose — I just forgot I used to have to. Six months ago, working with AI meant copying text into a chat window, uploading PDFs, dragging screenshots. Working inside one project folder at a time. It was pretty tedious and not really efficient, limiting the amount of work you can do with AI. The new biggest direction in AI now (and possibly in 2026) is when AI uses your computer to be able to do the same work as you. Now I'm watching AI read files across my entire machine, open applications, run terminal commands, and execute multi-step tasks across various files. Three companies shipped this in the same week. Manus launched a desktop app that runs on your local GPU and lets you assign tasks from your phone. Perplexity released what they're calling a "personal computer" — an always-on agent running 19 models on a Mac Mini (cloud based) for $200 a month. And Microsoft's Copilot is threading agents through Outlook, Teams, and Excel as a single coordinated layer. This solves a very big bottleneck for anyone using AI now. When AI can only see what you paste into it, you're the one deciding what context it gets, which files matter, what tools to use. I've been running desktop agents on real work for about 10 months now (through Claude Code) and now as OpenClaw pushed the demand into this direction, every major AI Company is focused on delivering similar capabilities. The change from working in AI chat-box to working with AI on your computer is probably a 5-10x productivity unlock. The issue most people face for now though (out of those who will even try it) is "it can do so much, but what do I do with it?", which is a typical AI problem for new adopters. The only way of solving this question, is trying to do the same manual things / repetitive workflows you were doing, and trying / trusting with AI to replicate the same. So far most of the clients I've met either don't know what AI can do, or afraid / live in a mindset that AI is not reliable and hallucinates too much (which is I would argue a 2024 / early 2025 mindset, meaning you're a year behind). Would be excited to hear if any of you are using AI on your computers today, and what are your top use cases!

Five 'AI that lives on your computer' products shipped within weeks... The chat-window era lasted about two years

@coingecko @Toriva_

Project that'll moon this month?

Complete weeks worth of work in minutes with RunMyBiz

First demo is here boys and girls! More coming soon. Run your business with AI employees

@MatthewBerman Exactly my thought! And using “local compute” for 24/7 agents

They’re building OpenClaw

One of the key issues with Perplexity computer, is that it has a cloud computer not connected to your machine and files, which significantly limited the possibilities. Moreover the usage included on it is very small and only on a $200 subscription, compared to Cowork that works on a $20 subscription pretty well. Manus on desktop is pretty new, but so far it seems like it’s a direct competition with Claude a Cowork. A few key differences - is that they also talk aboututilising your idle compute to run 24/7 agents, thus focusing on a more always on personal assistant you can chat with from anywhere and do things on your computer (like OpenClaw) vs Cowork focusing just on work live in chat

Anyone use Manus my computer / agents How does it compare to perplexity computer, OpenClaw and Claude Cowork?

@ManusAI Already installing

Today, we're taking Manus out of the cloud and putting it on your desktop. Introducing My Computer, the core feature of the new Manus Desktop app. It’s your AI agent, now on your local machine.

Sixteen thousand AI integrations exist right now. Slack, Salesforce, Notion, Google Calendar — you can plug AI into basically anything. That part is solved. Here's the part that isn't. I set up MCP (Anthropic's universal connector standard) for a client's sales team last month. After two weeks, a lot of people were struggling to properly use it. The team struggled to understand what it can and how it can do the work with it. And it's a pretty common problem in the industry (outside of our AI hyper adopters bubble). The conversation always starts with "connect AI to everything" or "AI can do almost anything", and stalls at "okay, now what do I actually tell it to do?" That's what Anthropic's Skills feature does. It lets you save reusable instructions that encode how YOUR team actually works. Not "you have access to Salesforce" but "when a deal hits Stage 3, pull the last five interactions, draft a summary, and flag open support tickets." Your judgment, baked into a repeatable workflow that anyone on your team can trigger. The companies getting value from AI are the ones that sat down and mapped their ten most common weekly workflows before they plugged anything in. What's interesting for me is - how many of you have actually used MCPs or Skills in your work with AI?

@trq212 I prefer “Agentic engineering” or “Agentic Building”

we need a better word than vibe coding man, Claude can create the most beautiful things