What if your company's Instagram was a neighborhood? Every post is a house. Every like is a car driving in. The more likes, the taller the house. You see in two seconds which post worked. Any data visualizations you'd recommend?

We'll be at AIBC Rome, 3 to 5 November, Fiera di Roma. We can still bring a few guests. DM me if you want to come.

Found ADHD traps this week. Can't stop watching them. Second look: how could we show company processes like this? City planners, Factorio and the polyrhythm music community have been at it for years. New rabbit hole. First try below.

𝗙𝗼𝗿 𝘄𝗼𝗿𝗸 𝗮𝗹𝗼𝗻𝗲 𝗜 𝗳𝗼𝗹𝗹𝗼𝘄 𝗮𝘁 𝗹𝗲𝗮𝘀𝘁 𝟮𝟬 𝗰𝗿𝗲𝗮𝘁𝗼𝗿𝘀. 𝗪𝗼𝘂𝗹𝗱 𝗹𝗼𝘃𝗲 𝘁𝗼 𝘄𝗮𝘁𝗰𝗵 𝗲𝘃𝗲𝗿𝘆 𝘃𝗶𝗱𝗲𝗼. 𝗜𝗳 𝗜 𝗷𝘂𝘀𝘁 𝗵𝗮𝗱 𝘁𝗵𝗲 𝘁𝗶𝗺𝗲... So we built a YouTube GPT. Add a channel or a single video, it pulls the transcript, and you can talk to it. What I use it for: Get summaries of content Chat with videos Pull out a strategy or playbook Save answers as markdown to our knowledge folder Repurpose for posts, blogs, newsletters Not for everyone. Helps me keep up. How do you keep up?

𝗘𝘃𝗲𝗿𝘆 𝗼𝗳𝗳𝗶𝗰𝗲 𝗵𝗮𝘀 𝗮 𝗳𝗼𝗿𝗺 𝗹𝗶𝗸𝗲 𝘁𝗵𝗶𝘀. Half the fields don't matter. Everyone fills it out anyway because someone set it up years ago and nobody questioned it. Revamp it or automate it.

𝗜𝗺𝗮𝗴𝗶𝗻𝗲 𝘁𝗼 𝗵𝗮𝘃𝗲 𝘆𝗼𝘂𝗿 𝗼𝘄𝗻 𝗽𝗿𝗼𝗺𝗼 𝗿𝗲𝗲𝗹 𝗳𝗮𝗰𝘁𝗼𝗿𝘆 We built a promo reel factory and we're using it for our launch. Most launch promos mean an agency brief, then waiting, then paying, then revising - we skipped all of that. Our agents did a great job: Ken handled marketing, Ben did design, Andy wrote copy - three AI agents on our team, not freelancers and not an agency waiting on a PDF. I'm building Haba - AI workspaces where agents help small teams to get work done. This promo is us using it on our own launch. DM me if you want to see how it works for your team.

𝗧𝗵𝗲 𝗯𝗶𝗴 𝗔𝗜 𝗰𝗼𝗺𝗽𝗮𝗻𝗶𝗲𝘀 𝗮𝗿𝗲 𝗿𝗮𝗰𝗶𝗻𝗴 𝗲𝗮𝗰𝗵 𝗼𝘁𝗵𝗲𝗿 𝗵𝗮𝗿𝗱 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄. 𝗪𝗵𝗲𝗻 𝘁𝗵𝗲𝘆 𝗳𝗶𝗴𝗵𝘁, 𝘆𝗼𝘂 𝘄𝗶𝗻. SpaceX just bought Cursor for $60 billion. Cursor already had the editor - where developers work. Composer 2.5 hangs with Opus on coding benchmarks, and it's cheaper. They were still bottlenecked on compute. Now they get Colossus to train the next model themselves. For the rest of us: better tools, faster releases, prices pushed down. You don't need to pick the winning lab. Build on the workflow layer and switch when something better ships. Messy for them. Good for everyone else. That's why we're building infra at Haba - so business teams get the workflow and context layer without waiting for the model war to settle.

𝗔 𝗰𝗵𝗮𝘁𝗯𝗼𝘁 𝗴𝗶𝘃𝗲𝘀 𝘆𝗼𝘂 𝘁𝗲𝘅𝘁. 𝗔𝗻 𝗮𝗴𝗲𝗻𝘁 𝗰𝗼𝗺𝗽𝗹𝗲𝘁𝗲𝘀 𝘁𝗮𝘀𝗸𝘀 𝗳𝗼𝗿 𝘆𝗼𝘂. Most founders I talk to think they're already using agents, but in reality they're in five chat windows, pasting brand context over and over, and copying output into their different software tools. An agent makes all this extra work redundant. It completes tasks in your stack - it handles files, notifies your team, gives you a result ready for you to sign off. Chatbots are fine for drafting. They're the wrong buy when you want to automate your work. Full breakdown: bit.ly/4gsTSy3 DM me if you want to see what the agent side looks like in a real setup.

𝗚𝗼𝗼𝗴𝗹𝗲 𝗔𝗜 𝘀𝗲𝗮𝗿𝗰𝗵 𝗰𝗵𝗮𝗻𝗴𝗲𝗱 𝗵𝗼𝘄 𝗰𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝘀 𝗳𝗶𝗻𝗱 𝘆𝗼𝘂. AI answers show up before the blue links now. Most studies put the overlap between page-one results and AI citations somewhere around 12–20%, so you can rank well and still not get named in the answer. Seer Interactive looked at 25 million impressions and found brands cited in AI search earned 35% more organic clicks and 91% more paid clicks than brands that ranked but weren't cited. Search your top five buyer questions in Google AI mode. If your brand isn't in the answer, that's your content list for next month.

𝗪𝗲 𝗮𝗱𝗱𝗲𝗱 𝗿𝗲𝗲𝗹 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝘁𝗼 𝗛𝗮𝗯𝗮. Video content in minutes, not a week with an agency. Drop your idea, or let our experts shape one for you. Haba builds the storyboard, reference pictures per scene, and outputs a finished reel. Coinbase-style demo attached — episode from a series we built called "The Pattern" (freelancer, wire delay, money to Cebu). Our demo, not their campaign. Storyboard to finished reel. DM me if you want to see how it works on your brand.

𝗟𝗲𝗮𝗱𝗲𝗿 𝗶𝗻 𝗔𝗜 — 𝗺𝗲 😳? Such an honor to be interviewed by Gabriel Borden from @AuthorityMgzine. Thank you for the feature. I hope I was able to provide value to everyone using AI to build their business or make it better. bit.ly/3PWaOSU

𝗔𝗿𝗲 𝘆𝗼𝘂 𝘀𝘁𝗶𝗹𝗹 𝘄𝗮𝘁𝗰𝗵𝗶𝗻𝗴 𝗬𝗼𝘂𝗧𝘂𝗯𝗲 𝗮𝘁 𝟭𝘅 𝘀𝗽𝗲𝗲𝗱? Too much great content out there that's directly relevant to my daily work. 2x is pretty much my default now and honestly, it still feels not fast enough. So here's one way to deal with all that great content: - YTranscript pulls the full transcript from any YouTube video - Save it as a markdown file - Paste it into your LLM, ask what you need to know - Or even better, take 10 md files and videos at once Want to go further? Automate it: - Every time you like a video or add it to a playlist, it goes into a spreadsheet - Every new row gets sent to your LLM - Configure what you need, summaries, key points, action items, whatever. I totally understand, this is not the right thing for everyone, but it's really helping me to keep up.

"𝗔𝗜-𝗳𝗶𝗿𝘀𝘁." 𝗖𝗮𝗻 𝘆𝗼𝘂 𝗲𝘅𝗽𝗹𝗮𝗶𝗻 𝘄𝗵𝗮𝘁 𝘁𝗵𝗮𝘁 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗺𝗲𝗮𝗻𝘀? Most people can't. Short version: AI-sometimes: you ask ChatGPT for ideas, copy-paste them somewhere. AI-first: the AI does the work, you steer, review, approve. The difference isn't using AI more, it's designing your work around it. It starts with structure - brand files, workflows, customer data. Without that, your AI is just a fast intern with no context.

Do you like chatbot interfaces? How do they make you feel? I'll go first: I hate them. Chat box. Wall of text. Scroll. That's 95% of AI products right now. There's a new term: Agentic UX. How humans should actually interact with AI agents. Building a good model is half the job. Making it feel right is the other half. The companies that fix this win.

𝗙𝗶𝗻𝗮𝗹𝗹𝘆, 𝗮 𝗹𝗶𝘁𝘁𝗹𝗲 𝗯𝗲𝗮𝗰𝗵 𝗵𝗼𝗹𝗶𝗱𝗮𝘆 𝘁𝗼 𝗴𝗲𝘁 𝘀𝗼𝗺𝗲 𝘄𝗼𝗿𝗸 𝗱𝗼𝗻𝗲. 𝗛𝗮𝗽𝗽𝘆 𝘄𝗲𝗲𝗸𝗲𝗻𝗱 𝗲𝘃𝗲𝗿𝘆𝗼𝗻𝗲.

𝗧𝗵𝗲 𝗼𝗻𝗹𝘆 𝘁𝗵𝗶𝗻𝗴 𝗶𝘁 𝘁𝗮𝗸𝗲𝘀 𝘁𝗼 𝗹𝗲𝗮𝗿𝗻 𝗔𝗜 𝗶𝘀 𝘆𝗼𝘂, 𝗮𝗻𝗱 𝗬𝗼𝘂𝗧𝘂𝗯𝗲. 𝗦𝗲𝗿𝗶𝗼𝘂𝘀𝗹𝘆. Here are 3 channels I highly recommend: Greg Isenberg — bit.ly/4rqgSAz Nate B Jones — bit.ly/46c1fUP The Boring Marketer — bit.ly/3OgAW9Z Don't act like a grandma. Don't complain that AI is taking away jobs. Truth is, better be open to something new, learning something new is awesome. Anyway, who am I to tell you? I bet you know already it's good for you. That's all, happy weekend.

Social media is drowning in content nobody wrote and nobody wants to read. That's not a prompting problem. It's people outsourcing their thinking. AI is a writing tool. A good one. But your job is to make it sound like you. Train it on your pre-AI writing. Emails, old posts, docs from before the tools showed up. That's your voice. Feed it in. Use the output as a draft. Edit until it's right. If you don't define your voice, the machine will. And it has no idea who you are.

𝗘𝘃𝗲𝗿𝘆𝗼𝗻𝗲'𝘀 𝗱𝗿𝗮𝘄𝗶𝗻𝗴 "𝗔𝗜 𝗔𝗴𝗲𝗻𝘁" 𝗼𝗻 𝗮 𝘄𝗵𝗶𝘁𝗲𝗯𝗼𝗮𝗿𝗱 𝗿𝗶𝗴𝗵𝘁 𝗻𝗼𝘄, 𝗯𝘂𝘁 𝗷𝘂𝘀𝘁 𝗱𝗿𝗮𝘄𝗶𝗻𝗴 𝗶𝘀 𝗻𝗼𝘁 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴. Making agents work in a real company is a different story. Here's what most people skip: An agent is only as good as what it knows about your business. You can wire up the best tools, the best framework, the smartest model - if the agent doesn't know your product, your customers, your pricing, your brand voice, it produces garbage. Garbage in, garbage out. That hasn't changed. So before you draw another box on a whiteboard, think about this: 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗰𝗼𝗻𝘁𝗲𝘅𝘁: your agent needs to know what your company actually does. Not a generic summary. Your docs, your positioning, your customer segments. The stuff that lives in people's heads and scattered Google Drives. If you don't feed that in, every output reads like it was written by someone who started yesterday. 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀: An agent that only answers questions is a chatbot. That's 2023. What actually moves the needle is an agent that sits inside your processes - marketing, sales, ops - and does the work step by step. Break the goal into tasks. Execute. Hand off. That's where it gets hard, and that's where it gets useful. 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗮𝗽𝗽𝗿𝗼𝘃𝗮𝗹𝘀: This is the one people talking too little about. You don't want an agent sending emails to your clients or publishing content without someone looking at it first. Every output needs a human checkpoint. Create, review, approve, done. Skip that and you're building a liability, not a product. We went from asking AI questions to giving agents responsibilities. That's a big shift. But responsibility without context is chaos. And responsibility without approval gates is dangerous. The framework and the model matter. But the real bottleneck isn't the tech. It's whether the agent actually knows your business - and whether a human signs off before anything goes out the door.

𝗢𝗼𝗽𝘀!… 𝗪𝗲 𝗗𝗶𝗱 𝗜𝘁 𝗔𝗴𝗮𝗶𝗻.𝗔𝗹𝗺𝗼𝘀𝘁 𝟭𝟬 𝘆𝗲𝗮𝗿𝘀 𝗮𝗳𝘁𝗲𝗿 𝗸𝗶𝗰𝗸𝗶𝗻𝗴 𝗼𝗳𝗳 𝘁𝗵𝗲 𝗗𝗮𝘀𝗵 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆 𝗶𝗻 𝗔𝘀𝗶𝗮, 𝘄𝗲'𝗿𝗲 𝗯𝗮𝗰𝗸. This time it's not fintech — it's Agentic AI for Business. We're launching the community right from the heart of Bangkok, starting with the Bangkok AI Meetup this Thursday. Tacos & Tech: Agentic AI Night. I'll be speaking alongside @jeancharlessalvin @✨-visouda-luangkhot-6b920857. Bring your questions - let us help you solve the challenges you're facing while trying to implement AI agents and automations for your business. If you're in Bangkok and building with AI (or trying to figure out where to start), come through. Thursday, Feb 19 6:30pm Missing Burro, near BTS Thonglor Link in comments or DM me. See you there.