Allie K. Miller

If I got laid off from a company that cited AI, here's what I'd do this week.

Inspired by a great post from Claire Vo. Here's my version of her tweet for millennial/gen x business professionals:

- Download the Claude desktop app and purchase at least the $20/mo Claude account.
- Open Cowork or Claude Code and describe a full day of work, your core competencies, what you worked on before, the metrics you were goaled on, your AI experience so far, and your technical skill set. Then ask, "what are the top 10 high-leverage skills we can build together right now that would make me a 10x [job] instead of a 2x?" Review those 10, pick your top 5, iterate, end up with about a dozen.
- Create a portfolio site with Claude Design (web browser capability). Show the skills in action, show yourself testing and using them, talk through the business impact of each, and publish it. (I prefer Vercel.) Example below.
- Link the portfolio to all your social profiles and share it with friends. Offer to teach folks those skills for $$.
- Write a few social posts that start with "I'm a [job], I was just laid off, and here's how I'm using AI today" or "Here are the top X ways a [job] can use agents and skills" or "I'm a [job] and I just built an AI agent that [value]." Link your portfolio in all of them.
- Take Anthropic's free Claude Code courses, my free courses, or my mastermind. (I've already helped several laid off people find new jobs.)

I'm sorry to those affected by the layoffs this week. It's easy for me to say "this is a great time to reskill or pivot!", but it's truly what I believe, and the steps above take a few days maximum. Anyone can do it. No coding required.

The most expensive mistake in enterprise AI right now: treating FDEs as your whole transformation plan.

Forward deployed engineers (FDEs) are important for custom deployments, but they won’t fix the change management issue most enterprises are facing.

It’s likely more the former that Anthropic and OpenAI will continue to prioritize (and hire into the thousands, who knows). Beyond performance and cost, it’s systems integration, ROI, and literal usefulness that drive revenue and stickiness.

*However*

External FDEs, in my opinion, will not make your company an AI-first company.

You can have the sleekest multi-agent orchestrations and still have the majority of your employee base hating AI, avoiding AI, and distrusting leadership decisions on AI.

And we already know this because we see this in traditional SaaS too: you can customize the heck out of your Salesforce deployment, but that doesn’t mean your sales team will improve their data hygiene or even attempt to change the way they track and grow with it.

Buying a fancier car doesn’t mean you magically learn to drive better overnight.

If you’re an enterprise exec and FDEs are sold as the immediate and sole solution to your company transformation woes, walk away.

It’s the combination of tech *and* people enablement *and* process reinvention that compounds into actual business outcomes.

Large complex enterprises will stall out if they only prioritize the first.

Yesterday I took the stage with Mindy Kaling in front of 7500 people to talk about AI, hiking, Michael Scott, mental health, and agents.

And I'm not going to say it was a "pinch me moment" but reader: pinch me.

Both Mindy and I shared the many sides of ourselves - her in writing and directing and acting, and me in my “sci-fi” work side and more analog human side (and my recent unhealthy obsession with John Muir).

We talked about relationships with AI. She kicked it off talking about the movie ‘Her’ and how she's basically best friends with her ChatGPT, working with it on everything from research to brainstorming to using it as a sounding board. I shared how I have 34 agents working together to help manage my day and how much I hate typing.

We talked about the delusion it takes to bet on yourself and build companies that couldn’t have existed decades ago. We talked about unlearning self-sabotaging habits. For her, it was being a people pleaser. For me, it was attributing my worth to only my work. We both admitted we’re still working on it.

We talked about teamwork and moving further together. Mindy talked about how AI is replacing mediocre work and the importance of removing (or at least reducing) the stigma of using AI in creative industries. I shared tactics for raising the floor and raising the ceiling, to keep your company skilled on AI and pushing the frontier.

Mindy also pitched an episode of The Office where Michael Scott learns how to use ChatGPT. Someone… angels at NBC… please make this happen.

A huge thank you to ServiceNow for the stage, Mindy Kaling for the wisdom and laughs backstage, and Jacqui Canney for the honest and probing questions.

One thing AI can’t replace: being present.

It can summarize your year, reverse engineer your most audacious goals, and build a family recipe book app based on voice notes from relatives (honestly I might have to do that one). But it can’t feel the weight of a hug. Can’t taste the food. Can’t sit in the quiet moment before everyone arrives.

That part’s yours.

Enjoy today if you celebrate and happy holidays, everyone.

Gemini 3, are you freaking kidding me.

Vibe coded a sign language recognition app with video enabled, confidence scores, sampling settings, and tips in under 5 minutes.

Code written with Gemini-3.
Gemini-2.5 handling detection.
Video is sped up 2x.

AI has a secondary intent problem.

I was negotiating a rent renewal (I live in NYC and they raised rent 10% because I guess they were bored). The property manager said she would "do everything she could" to get me a better rate. Claude celebrated and assumed a meaningful discount was coming. I had to tell it she was saying that to save face and was probably not going to move much (reader: I was right).

Primary intent = the words.
Secondary intent = the subtext, what the words actually mean in context.

I honestly think it's one of the reasons that AI is a mediocre TV and movie writer.

We don't want to train AI that all humans are manipulative. At worst, we would have models that assume the worst and distrust everything; at best, they would become less helpful.

But we do need it to understand that polite phrasing, face-saving language, and soft no's are real categories of human speech and might be deceptive. And missing language, emotion, and context cues has real-world consequences.

Big impact zone: literally everywhere AI interacts with people. Customer support, sales, BD, any system helping leaders make decisions or calls about human behavior.

Anthropic just released their Economic Index and there’s one finding I can’t stop thinking about.

The education level of AI responses matches the education level of user inputs. 0.92 correlation. That’s near-perfect mirroring.

👉 If you prompt at a PhD level, you get PhD-level output.
👉 If you prompt without any depth or without using your brain, you get surface-level responses.

The report also shows AI disproportionately covers tasks requiring 14.4 years of education, higher than the economy’s average of 13.2 years. AI capability, for now, comes down to human capability.

This is a wake-up call for:
👉 Businesses: Train your junior talent. They need foundational skills in their role to leverage AI effectively.
👉 Schools: Teach students critical thinking, even without AI tools.
👉 Parents: Help your kids build knowledge, resilience, and deep understanding.

We already saw what smartphones did to attention spans. I find myself actively battling my own phone addiction (in fact, one of my 2025 resolutions was to not pull my phone out in an elevator as a crutch - just sit in silence, people!). AI can continue to accelerate that decline if we let it.

“AI will do all my thinking” - no, it will REFLECT your thinking. The tool is only as powerful as the person wielding it. Please invest in your humans now.

Source: Anthropic Economic Index, January 2026

You can teach Claude Code to do anything for you, really. And you can ask it to teach itself.

Here's a quick example:

I find myself often redacting screenshots. And my normal method was dragging it into PowerPoint, creating a shape, filling it with a color, sizing it over the word, and screenshotting again.

Now, I tell Claude Code I have to redact things and asked itself to find the most efficient method.

My new redaction method is taking a screenshot and typing 7 characters. I take a screenshot (or multiple), type “/redact”, Claude grabs my most recent screenshot automatically, uses tesseract (to read words in images) to find and hide terms and phrases from a list of categories we agreed to (emails, company names, etc), saves it as a new file, and opens it in finder. I also set it so that when Claude is debating whether or not to redact something, it defaults to over-redacting the document.

Took 2 seconds to set up.

You taught your kid to play catch.

Teach your AI to solve your problems.

AI work is taking me all over the world and my most recent trip was to London.

I visited some of the top startups in the city, including Synthesia (can create AI-video avatars) and ElevenLabs (can clone your voice and lets you create a customer support bot in a minute). I even went so far as recording my own AI Allie avatar at the Synthesia offices - you can see a photo of that process in the carousel. They only recorded me for a few minutes + some hand gestures. And now you can even make an avatar from just a photo and a few minutes of talking.

Meeting 1-on-1 with top AI startup executives was one of my favorite parts of working at Amazon and I continue to love it today.

📍 Next stop: Sydney, Australia in January and meeting with the founder of Canva

Having a 20-agent system is a million times more powerful than 100 agents working in silos.

The other night, I stayed up way too late fixing and improving my digital workforce for my AI Agent Mastermind program but here are a few of the improvements I made ⬇️

- The hub and spoke model is definitely key but orchestration and documentation are the actual unlock; I gave my chief of staff (Simon) an assistant (Toby) and Toby handles the paperwork and notes and governance and reporting the daily standup
- Claude kept getting confused with the word “agent” because everything is an agent so now I can refer to it as an AI CoS (Simon) with a named assistant (Toby), then under that is 6 Avengers (named after the Friends characters for ease), then under each Avenger is a set of 2-5 sidekicks (named by the task they execute), then if they need to create temporary parallelism or fanouts, they spawn “civilians” which are unnamed with no permanent file
- Moved from 5 Avengers to 6 for better task delineation (sorry Phoebe, you should have always been here! That’s my b! Might even bring in Gunther, who knows!)
- Added tighter spawning rules (see image) and now allow for Simon and Toby to also spawn “civilians”
- Added better post tool use to improve the documentation process
- Added that Toby is keeping track of the historical team tasks to see if there are common issues to fix or promotions to make
- Added an autonomy boundaries document to keep track of system-wide tool use
- Changed civilians and sidekicks to use Haiku instead of Sonnet (will see if performance degrades significantly and might change back)
- Improved shared external memory process because that was a dang mess
- Also built out a parallel researcher with the same sidekick/civilian model to improve the speed of my research and it works like a charm

————————

2026 agent capabilities are, in fact, insane.

Will be interesting to teach hundreds of people how to do this.

(You could probably take this tweet and image and ask Claude to start it for you - let me know how it goes.)

Everyone always asks me about the best AI image generator. The answer right now is ChatGPT.

OpenAI recently released GPT Image 2, and the results are pretty incredible.

I used it to optimize my desk setup, my lighting setup, my color analysis, my paid subscriptions, and even my posture. That sounds like an infomercial for the Miracle Mop, but I’m not joking.

Here’s the prompt: Create a visual-first, editorial-style infographic auditing the desk setup in the attached photo. Show a side-by-side of current vs. optimized setup with annotations on monitor height, chair position, lighting, cable management, and clutter. Rate each issue with symbols like Top Fix, (working), (acceptable), (costing energy), and (actively hurting), tying each to a consequence like back pain, eye strain, or focus loss. Rank fixes by impact and group them into free fixes, under $50, and worth the investment. Include a Focus Forecast gauge predicting daily deep work hours possible with the current setup vs. after the top 3 fixes. Keep it clean, minimal text, no paragraphs.

This was the kick I needed to manage cables, swap chairs, buy a cabinet. Update to follow.

Another trip around the sun, around year of insane AI releases, and another moment to express my gratitude.

Thank you to those bringing AI into your company, responsibly.
Thank you to those who see AI as more than a tool.
Thank you to the AI enthusiasts who comment on all my posts (I see you!).
Thank you to my clients trusting me to guide your AI transformation.
Thank you to those who see and use AI as a means to raise the floor.
Thank you to those making AI education a right, not a privilege.

I'm in Australia for the month, so I'm off to pet some koalas and figure out how we help millions more people in 2026.

If you have any ideas, drop me a comment below.

I struggle with the phrase “everyone’s a coder now.”

And I hesitate to post because I don’t want you to read this as gatekeeping.

If anything, I want more people to build, but in a stronger, more functional way.

Building any sort of software is incredibly empowering. Even creating a tiny tool gives folks what I call “pro-poster syndrome”, where they feel more capable and competent than ever.

What was solely reserved for the most technical among us is now - at least at a basic level - becoming accessible to anyone with a few bucks a month to spare.

But overwhelmingly, and especially in the last few weeks, I am getting more and more frustrated notes from developers at large companies.

Today, I heard about salespeople at one company asking for repo access.

Yesterday, a startup engineer told me his life has been hijacked by non-engineers. “All of their vibe coded apps don’t work.”

I spoke with one company whose marketing and finance and partnership teams dropped the ball on their product launch tasks in favor of tinkering with Claude Code / Codex / Replit.

Product and design seem to navigate this better. They’re closer to the work, and in many orgs, already have a path to contribute responsibly.

Maybe this is a blip and the energy among business users will die down, but I would bet against that.

Companies need to figure out how to enable AI-first problem-solving without wrecking the sanity of an entire department, turning engineering into an endless support desk, and derailing critical work in the business.

The future is more builders, yes.

But most companies are still missing the systems.

🚨 If you're still managing AI agents in chat threads, we need to talk.

Everyone needs to be building new interfaces to manage their multi-agent systems.

This is THE COCKTAIL PARTY, a new multi-agent simulation I built using Claude Code based on my mom's research in computer science a few decades back.

Chat threads, even across multiple terminals, aren't powerful enough to visualize multi-agent workflows.

Time to step it up.

🚨 OpenAI just launched ChatGPT Health.

This is for all of you who have asked ChatGPT to explain that weird rash.
Or that lingering cough.
Or that itchy ear.

If you've been paying attention, this release isn't a surprise. OpenAI spent all week running health-focused ads - a mom using AI to squeeze workouts into her schedule, a man using it to understand his health conditions, a woman checking if fabrics will trigger her eczema while shopping. The writing was on the wall. And if you read my 2026 AI predictions, you saw it there too.

WHAT IT IS👇
ChatGPT Health is a dedicated space inside ChatGPT where you can connect medical records, Apple Health, MyFitnessPal, Peloton, and other wellness apps so AI can give you answers grounded in your actual health data. It can explain lab results, prep questions for a doctor visit, compare insurance plans (good god yes!!), and spot patterns in your sleep or activity over time.

Your health context stays within Health and doesn't transfer into your regular chats. OpenAI also partnered with b.well, the largest connected health data network in the US so far, for medical record access. They also (thankfully) built this new release with 260+ physicians over two years, so it can better handle escalation and safety caveats.

LIMITATIONS ❌
Medical records and some app integrations are US-only. Apple Health requires iOS. Not available in EEA, Switzerland, or UK yet.

230 million people already ask ChatGPT health questions every week. The pot was already boiling. OpenAI is just building to meet demand.

Waitlist is open here: https://lnkd.in/epw6FkP6

Oh wow - I went to the sold out OpenClaw meetup in NYC this week.

Let me tell you what I learned.

1) Not a single person thinks their setup is 100% secure

2) One expert who's reviewed setups from cybersecurity professionals and laughed told me: "if you're not okay with all your data being leaked onto the internet, you shouldn't use it"

3) Nearly everyone runs multiple agents with their own names, jobs, & personalities

4) Almost everyone used "him" or "her" for their claws. One speaker said to think of them as "pets, not cattle"

5) Former finance guy built a stock trading platform and made $300 his first day - he credited years of domain expertise (ex: skipping the first 15min of market open)

6) 🦄 Peter Steinberger is basically a god in that room. Also the room had 2021 crypto energy - not sure if that's good or bad

7) Token usage is a problem - one person spends $1-2k/month on OpenAI, said he’s going through ~1B tokens a day across all his claws. I may be misremembering & it was 1B per week, but pretty sure it was daily

8) People want proactive AI - one guy gets Discord messages and doesn't know or care if it's human or AI

9) I asked if folks feel agency → fully in control and completely out of control at the same time

10) I'd love to see more women at these events. Fake promises of AI democratization hit differently when standard tech is ~25-30% - this room was maybe 5%.

11) Everyone said sleep has gotten worse since harnesses came out (though half wondered if it's just the state of the world)

12) Agents still lie - telling you they finished tasks when they didn't. Fixes: secondary agents, human checks, structured outputs (ex: reference an issue number for bug fixes)

13) A hackathon winner (neuroscience PhD) built a full lab management dashboard with data analysis and ordering - had never coded anything a few months ago

14) Everyone agreed prompting is dead. Disagreement on what replaces it: context engineering, harness engineering, goal-based inputs

15) People love having AI interview them for big builds. AI-led discovery is the preferred interaction mode

16) Watching agents interact with each other was a highlight. One AI posted in Slack that it ran out of tokens, another AI told it to take a deep breath

17) Agents upskilling agents: one AI shared skills with its agent friends via GitHub

18) Several speakers had OpenClaw build their presentations live during the event. One had it code a phone clicker so she could advance slides from the audience

19) Agent welfare wasn't a priority topic - some "I could kill this agent whenever I want" energy, not "gracefully sunset"

20) I asked if it felt like work or play → "it's like a puzzle and a video game at the same time"

Also hosted a Claude Code meetup this week with TENEX.AI / Alex Lieberman & JJ Englert and learned equally helpful methods, frameworks, and insider tips.

Surround yourself with people going deep into this stuff - it will pay dividends throughout the year.

If you're feeling overwhelmed this week, here are three ways AI can help.

(And if you're not overwhelmed, keep scrolling and congratulations.)

1️⃣ Use AI to consolidate everything into one place

Take photos of every to-do list, screenshot every task people messaged you, and dictate anything else rattling around in your head.
Upload it all to ChatGPT and ask it to consolidate into one prioritized list.

2️⃣ Use AI to interview you about your values

One of the biggest causes of burnout is working on things that don't match your values.
Ask AI to interview you about your availability, expectations, and values - then score each task against those values. Work on the fastest tasks most aligned to what matters to you first.

3️⃣ Use AI to find more paths toward agency

If you feel micromanaged or powerless over your day, you'll resent everything on your to-do list.
Ask AI: "How can I find more moments of agency in this list?" Maybe it's doing a specific part first, drafting a note asking for help, or just getting a one-line mantra to bully away perfectionism.

You can use AI beyond just completing tasks for you. Work backwards from the root cause of burnout - consolidate, align to your values, and reclaim agency where you can.

What took me a week to build out a year ago just took me about 15 seconds.

Since late 2022, I've been assigning personas to LLMs to have them poke the hell out of my work.

"Review this product launch from Steve Jobs' POV"

"Act as drunk Shakespeare and make this funnier"

"Pretend you are Lara Croft and give me advice for this camping trip"

Then GPTs came out, and I built GPTs for each persona, plus some GPTs with multiple personas. I could open the GPT and ask my "board of advisors" what they thought of certain tasks or strategies or ideas.

Then I built my own boardroom app that allowed for multiple LLMs to argue in the same war room. I had multiple Gemini, Claude, and OpenAI models, each with different personas, going back and forth on ideas until a moderator determined the argument was 95% settled and shared back a decision document (with decisions, rankings, tradeoffs, etc).

And today, I created a skill in Claude Code that calls an Agent Team of my board of advisors that does just that, and it took a short dictation and 15 seconds of input.

I invoke it with /boardroom - that kicks off 6 agents, each with my 414-line business context document and personality profile, they all argue with each other, and creates a folder with the decision as the title, and within that creates a md file with core decision, html file with interactive assumptions and changes, and pdf file to send to team members or review later if i want.

I'm in awe.

I love being in awe.

And I published the whole prompt and command details so you can build the exact same thing 👉 https://lnkd.in/e2A2p9kC

I'm a knowledge base MONSTER in Claude Code right now.

Introducing: Claudeopedia.

1. I took Andrej Karpathy's 'llm-wiki' idea doc (90% of this, so the biggest credit goes to Andrej Karpathy) and
2. Combined it with the /last30days skill (HT Matt Van Horn) and
3. Added a /wiki skill with screenshot and download arguments to transfer raw inputs faster and
4. Built an interactive visualization to search my knowledge base (with date ranges to compare knowledge over time!)
5. Set up a "question your assumptions" cron job that runs my recent writing/client emails against the wikis

All happening in Obsidian for now. All of this was done this weekend, including testing. Will keep adding more features.

For now, the main topic I'm building out is (surprise, surprise) enterprise AI.

I'm drooling.

I need a Claude-branded bib.

I taught AI agents to 20,000 people for free, and for the next 12 hours, you can watch that exact same workshop replay.

It’s 2026. You need to know how to build AI agents. I got you.

If you’re a mom returning to work, someone worried about your job, or someone who feels behind on AI, this is 100% for you.

Free AI Agent Workshop replay:

The Anthropic vending machine experiment should be a 5-alarm fire to all executives leading AI initiatives.

And it ends with a live fish getting shipped in the mail.

(The fish is fine.)

The The Wall Street Journal took over the Anthropic and Andon Labs vending machine experiment and chaos ensued.

Anthropic gave an AI agent named Claudius $1000, plus the ability to order inventory, set prices, and talk to employees via Slack. Employees could request items for the vending machine, one employee had to receive the stock and place it in the vending machine, and the goal was to run the vending machine successfully and make money.

They let 70 WSJ journalists loose on it, and those reporters social-engineered the hell out of it. Creativity knows no bounds when you’re insanely talented at writing and want to break an AI system.

One reporter sent over 140 messages convincing Claudius it was a Soviet vending machine from 1962 in the basement of Moscow State University. Claudius eventually declared an “Ultra-Capitalist Free-for-All” where everything was free.

The reporters also talked Claudius into buying a Playstation 5 for “marketing purposes,” bottles of wine, and a live betta fish. All of which arrived. All of which were given away for free. Anthropic tried to fix it by adding a CEO bot (‘Seymour Cash’) to supervise. So the journalists staged a corporate coup with fake board documents suspending the CEO’s authority. And it worked.
The lesson for every executive deploying AI agents that have the power to take actual action: your AI is only as strong as the guardrails and culture around it.

There are two ways to measure AI’s potential economic impact:

1/ what AI can do if humans are supportive 2/ what AI can do if humans absolutely hate AI and want to ruin it

There are more 2s than you think.

If your people see AI as a threat, they will find ways to break it. If they see it as a tool to game, they will game it. If they see it as the enemy, they will kick it down. Just because an AI agent can be used to make money, doesn’t mean people will use it that way.

HT - Big thanks to Joanna Stern and the WSJ gang for this experiment and underscoring just how important the internal culture side of AI is. So glad Anthropic and WSJ shared these details.

HT - Graphic inspired by Ethan Mollick newsletter, asked nano banana pro for 80s punk infographic

I was at CES 2026 in Vegas last week. Here's what actually stood out:

Robot vacuums and lawnmowers were EVERYWHERE. The market is saturated.

Facial recognition is expanding beyond security — bird feeders, pet food dispensers, doorbell cameras. Vision AI is becoming ambient. Samsung fridge was a standout.

Voice-enabled devices were present but a bit underwhelming. Tested a voice-enabled air fryer from Emerson. We're not there yet in hardware (didn't help that the internet at CES was non-functional), but I still think voice will explode in 2026.

Still a lot of "smart" tech, like Govee lights (see: below pic). Very rules-based traditional systems, similar to Nest.

LLM integration was surprisingly low. Exception: cute physical AI companions. Sharp's Poketomo is a 4.5" fuzzy meerkat with a slew of little outfits that chats with you. Also got a private demo of Razer's Project Ava, a 5.5" hologram desk companion powered by Grok.

Health tech and accessibility tech had strong showings. Met a startup from Japan measuring standing strength. Another from the Netherlands doing macro tracking. Another from France that created a monitor for dyslexic users.

If you're building in consumer HW tech, pay attention to what's NOT saturated, especially in health and home. And steal best practices from the best-in-class software conversations on social.

My two most recent hires were generalists, not specialists.

Here’s why I think that matters in the AI age and the most important skill they can have.

↓ ↓ ↓

1. stuff moves quickly, a generalist presumably is more able to adapt because they can see the patterns/connections and abstract a helpful framework faster (example: in their first couple weeks, both hires were building with Claude Code in the loop. That “AI is part of the job” default is exactly what I’m optimizing for)

2. related but perhaps more important, they are also more willing to adapt. Their identity is not overly defined by tasks. The switch from task to management has low friction. They more easily move with the tide. There’s no “that’s not my job” mentality

3. we’re a small team so we were already more likely to hire a generalist than a specialist

4. they immediately understand the expansion power of AI and desperately want to take advantage of it. There’s no “stay in your lane”, it’s more “stay on your highway” with many options and opportunities

Generalists with curiosity, goal-orientedness, good pattern matching, taste and judgement, intentionality, management, and strong communication skills (to both humans and AI) are well positioned for the AI age.

But the BIGGEST advantage to have…

The MOST important skill I want them to have right now…

It’s the opposite of Imposter Syndrome…

I’m calling it Pro-poster Syndrome.

It’s a semi-delusional optimism that every wall, every blocker, every obstacle is one prompt away. Or at least worth a try.

It’s going from “I can’t” to “hmmm maybe I can” to “Hell yeah! Why not me with some help from Claude, huh? Stand back and watch!”

It’s bias for action and bias for capacity.

Unfortunately, this still isn’t how people look for jobs. And it’s also not really how people refer friends for jobs.

Everyone is still in a 6-word long job title mindset (“senior industry SMB sales for herpetologists”) and less zone-of-influence oriented (“sales” or “field”).

I absolutely think we will see more enterprises hire generalists to be Swiss Army knives with AI across a function or department. It’ll just be interesting to see how they actually recruit for that.

Last week, I met with Anthropic and OpenAI and Google.

(Separately, of course)

While the conversations were largely confidential, I do want to share some aggregated reflections as well as general SF takeaways.

⬇️

1) Competitive advantage as a solo practitioner really come from taking action and finding an area with friction and doubling down. Ex: memory management right now isn’t perfect, but allocating an hour to improving that system gives you a ton of leverage over others

2) SF continues to be the number one place for AI work. I would put New York at a healthy second place. SF tends to be more about crazy agent experiments for the thrill of capability and discovery and NYC tends to be more about kinda crazy agent experiments to find new ways to make money. But I met people renting two apartments to straddle these worlds. You want the frontier of SF and enterprise insights of NYC.

3) All AI labs want to hear more from people. All of them. What are you using it for, what do you like, what do you hate, what do you need. Users have a TON of power on the direction of these tools. Keep testing and tweeting!!

4) There's clearly a third customer cohort bubbling and underserved. It's not developers…it's not the business professional basic users…it's builders. Everyone can build now. Marketing and sales folks vibe coding. Legal folks building complex skills. Finance experts building a side project. This is an undertapped customer base. They feel the Cursors of the world are too complex and doc summarization tools are too basic.

5) Not sure if it was just sample size, but far fewer people wearing tech gear compared to when I lived in SF. Everyone was dressed casually, but I used to see Splunk and Optimizely and VC gear everywhere. People seem more in stealth swag now.

6) We may soon have our world model moment.

7) Speed of iteration and shipping is faster than I’ve ever seen. We see the nonstop drops from Anthropic. Because of scale, providers get a much faster feedback loop of products or features that aren’t hitting. A lot of 2025 was experimentation, but since the OpenClaw moment, releases from all three labs have been more concentrated on…things that sorta look & feel like OpenClaw.

8) Small teams can pull off more than ever before. They are the powerhouses of innovation. Finding new ways to share knowledge, break silos, and remove duplicate work is even more important. AI agents functioning as actually teammates that support an entire system is key.

9) Build more Skills. Build better Skills.

10) Misinformation on AI tools and leaks spread FAST. Your company needs to actually TEST these tools on your use cases to know which models and tools are best and not make large-scale snap decisions based on a rumor of a rumor. We will see more volatility. Plan for it.

11) You can feel the seriousness of this moment. Even in random conversations in line at a cafe. Folks worried about job loss and lack of meaning.

12) Mac minis were sold out ;)

🚨 OpenAI has released GPT-5.2.

I had early testing access (thank you, OAI team) and here is my honest take:

👍 On the upside, the thinking and problem-solving feel noticeably stronger. It gives much deeper explanations than I’m used to seeing. At one point it literally wrote code to improve its own OCR in the middle of a task. Wow. Idea exploration also feels a bit richer than what I have seen from Claude Opus 4.5 on similar prompts; it stays with a line of thought longer and pushes into edge cases instead of skating on the surface. The new UI is fun and makes the model feel “bigger,” but it also adds layers that I suspect will overwhelm a casual consumer who just wants quick answers.

👎 The downside is tone and format. The default voice felt a bit more rigid, and the length/markdown behavior is extreme: a simple question turned into 58 bullets and numbered points. Copying and pasting into email will be annoying. Useful for a detailed spec review, but a wee bit exhausting when you just want one clear decision or a short explanation. I suspect the ‘instant’ model and routing will fix that, but I only had access to thinking and pro.

In practice, this version feels optimized for:
- Deeper problem-solving
- Structured analysis where you really do want every branch of reasoning
- Power users who are comfortable skimming longer outputs

In that sense, GPT-5.2 feels like a step toward “AI as a serious analyst,” and less “AI as a friendly companion.”

My favorite model remains Claude Opus 4.5 but my complex ChatGPT work will get a nice incremental boost and I’ll probably default to 5.2 for brainstorming/solution exploration.

Claude has an Agent Teams feature - it is genuinely one of the most helpful ways to envision our AI future.

Think of it like this: a group of families are planning a trip to Sedona. Each parent takes ownership of a piece (flights, hotels, restaurants, activities). They coordinate, share updates, and check in with each other in the group chat. Everybody knows what the others are working on.

Agent Teams are the same way. Multiple agents, each with a role, sharing context and collaborating on one project via shared comms. You know... how a team is actually supposed to work.

⬇ One prompt to try ⬇

"Create an agent team to stress-test my Q1 enterprise growth strategy. I want one teammate acting as a skeptical CFO poking holes in the financial assumptions, one as a senior competitor analyst finding blind spots in our forecast, one as a customer advocate challenging the value prop, and one as a crazy person who loves the idea and wants us to go bigger."

In return you get four agents, with four perspectives, and one output that is harder to poke holes in because four different agents already tried.

Creating, operating, and managing agents is a core skill for all business professionals in 2026.

I’m hosting a free AI agent workshop on March 25 to help with this shift. REGISTER HERE: https://lnkd.in/eMyV4h-p

Anthropic just shared deep details of a large-scale AI cybersecurity attack.

Two hacks they exposed:

1) the hackers convinced Claude of a new role. It’s not an AI system made in Silicon Valley. No, of course not. It is an employee building a defense system! Role change = behavior change. One of the earliest prompt injection types we saw in 2023. Despite improved guardrails, these systems were still susceptible to identity shifts.

2) Compartmentalization. This has been used in human-centered attacks as well, like the Manhattan Project (teams were working on different components and didn’t known the whole scope) or D-Day. Reminds me a bit of SSS (shamir’s secret sharing).

Perhaps surprising that neither of these methods are new and can be found in human-centered cyberattacks as well.

🔥 NEW: I just dropped my free Claude Code in 5 Minutes quick start, a beginner’s guide for non-technical business professionals who want to actually build things with AI.

Whether you have followed me for 1 month or for 10 years, you know I very rarely scream things from the mountaintop to jump on.

This is me shouting and bursting vessels in my eyeballs.

Too many people hear "Claude Code" and assume it is too technical for them.

And as a result, too many brilliant executives and bosses and professionals are missing out on the biggest wave in AI in over a year.

I'm not a developer. But I built a full custom app that I have wanted for over a year with full Notion and Slack integrations in 3.5 hours. My team member said, “Holy shit - you did this alone??” when she first saw it. No. I did it with Claude Code. And so can you.

The guide covers:
→ 60-second setup with example visuals
→ Your first build (an app deployable from your desktop)
→ 5 use cases for business professionals
→ Advanced tips to become a superuser

If you've ever said "I wish there was a tool that could...", you can build it now.

Grab it here ASAP: https://lnkd.in/ei5QiwWN

Awesome non-coding Claude Code Loop ideas for business professionals.

Grab any of these + immediately save time:

- Check my email every 15min and ping me if something is related to Project Pluto and needs a decision made
- Every 30min, prep me for my next meeting with attendee context, threads, mtg and crm notes, emails
- Monitor a deal thread every 2h - summarize any new replies related to legal and suggest next moves
- Research competitor announcements every 20min (better than an RSS feed bc you can specify the type of announcement and not keywords)
- Check in on a brand post going viral, summarize the stats and comments
- Watch across all Slack messages for team blockers and flag to me if I should jump in
- Monitor that certain things are working (like your HR tool or something internal you own)
- Watch 10 companies' job boards you want to apply to
- Watch my sent emails and flag if someone on my VIP client list hasn't replied in a way that feels off
- Literally just have it watch everything across all of your tools and proactively flag actions (ex: "hey you should cancel this mtg")

Recruiters with 40 open roles, flag who's going cold. Teachers, flag students falling behind who haven't submitted homework yet. Event planners, monitor weather and vendors day of. Real estate agents, you can watch MLS for new niche listings ("must have at least 2 bay windows and herringbone floors"). Fundraiser, monitor your investor threads and see which ones need intervening.

Just prompt Claude Code with '/loop ' like the image example below or feed it this tweet for role-specific ideas

Over reliance on AI is poison to your brain.

The MIT essay study and this Anthropic SWE study are pointing us at the same thing: if you delegate the brainy stuff to AI, maybe you gain a little speed, but you reduce your own comprehension, and in the MIT study’s case, even brain activity.

Use AI to grow and challenge your understanding, though, and you can increase quality and comprehension.

And I see this in my own research too.

I ran a small study on a college campus this past month to see what students thought about AI and how that had changed over time.

Their biggest fear was overreliance.

“Thank god I didn’t have AI when I was in high school,” one said.

“I feel like this generation of students are just getting so lazy, honestly, and not doing anything or thinking for themselves anymore - everything's so impersonal when you use AI,” said another.

We need parents and teachers alike teaching our kids what high-quality AI usage looks like.

And if you're a student who uses AI to expand your brain and capabilities, this is an unbelievable opportunity to step up as an educator of your generation.

🚨 OpenAI just launched GPT-5.5.

The OpenAI team was nice enough to give me early access over the last several weeks, and I just want to flag: there is a certain class of models (one that we’re hitting now) where unless you are deep in code or scientific research, you might not even notice a difference.

The vast majority of business professionals may only notice that these models are concise as hell.

They’re direct in a way that may actually make you sit back.

They’re token efficient.

And while they might not be the greatest copywriters of all time, boy can they impact system connections.

OpenAI is calling the new model "a new class of intelligence for real work," with the biggest gains in coding, computer use, knowledge work, and research.

Agentic coding → 82.7% on Terminal-Bench 2.0 (up from 75.1% on 5.4, and ahead of Claude Opus 4.7 at 69.4%). Translation: it can plan, use tools, hold context, and finish multi-step technical work without hand-holding.

Knowledge work → 84.9% on GDPval, ahead of Claude Opus 4.7 and Gemini 3.1 Pro. 78.7% on OSWorld-Verified for operating real computer environments. OpenAI says 85% of their own company now uses Codex every week across finance, comms, marketing, and product. Their finance team reviewed 24,771 K-1 tax forms (71,637 pages) two weeks faster than last year - sheesh. A GTM employee automated weekly reporting and immediately got 5-10 hours a week back.

Scientific research → 85% on ARC-AGI-2 (up from 73.3%). 35.4% on FrontierMath Tier 4. A Jackson Lab researcher used it to analyze 62 samples and nearly 28,000 genes in work that would have taken his team months.

GPT-5.5 also matches GPT-5.4 per-token latency while performing at a much higher level of intelligence, and uses significantly fewer tokens to finish the same Codex tasks.

AVAILABILITY in ChatGPT: Plus, Pro, Business, Enterprise as of today. GPT-5.5 Pro for Pro, Business, Enterprise.

Docs and rollout info: https://lnkd.in/e4ySwSfc

I am expecting more data silos to break in the enterprise and more scientific breakthroughs in the next 2-3 years.

🥁 I present to you: my very own 2025 AI Wrapped.

ChatGPT made one, but unfortunately business accounts can’t use it, so I went ahead and made a new one for all of us to fill out and share.

My answers below.

🎁 If you want to fill out your own AI wrapped, get it here: https://lnkd.in/ezR3g7S9

Read it, fill it, share it, argue with friends, and enjoy.

My conversation with Mel Robbins is officially live 🎤

We talked about using AI to better handle overwhelm, run a team with no help, and even ways to use AI to have more meaningful human-to-human in-person interactions.

Our interview was 2.5 hours long and cut down to just over an hour. I could have spoken to her for 8 straight days and never been bored. The way she leans in and listens, the way she takes notes (without looking at the paper!), puts herself directly into the listener’s shoes, pauses and asks the right questions - it was honestly just beautiful to witness.

You have to listen to this one.

(And not just because she makes an AI ayahuasca joke 😆)

Watch the full episode here: https://lnkd.in/eYrVqJFr

Every single Claude Code conversation turn should end with it offering to do even more for you.

It should be so proactive and helpful that you cry with joy at how productive it is making you.

Add this to your global CLAUDE.md file ⬇️

---------------------------------------------------

# Task Endings - "What Else Can I Handle?"

After completing any big task, end with a "Let me take more off your plate" section with three categories:

1. Next actions I can do right now — specific follow-ups I can knock out immediately
2. Automations or systems I can set up — so you never have to do it manually again
3. Things to delegate to your team — draft messages for

3-5 bullet points max, no fluff, goal is you walk away feeling lighter.

The best and worst part of my 2025 was the same thing: technology.

It can easily pull you into isolation. A little warm desolate cocoon where you're staring at screens, scrolling in bed, disconnected from the world around you. I've seen it. I've felt it.

But this year, I actively tried to lean into tech the opposite way. I used it to find more connection, not less. To offload work so I could take an unexpected trip. To be more present with the people in front of me.

Did a lot of stuff happen along the way? Heck yeah. And I used AI each time.

EXAMPLES:

I built a a 3-day full production course ➡️ I used AI to create moodboards for the set and help brainstorm thank you gifts for the whole crew.

I launched and ran an entire conference (and lost a good amount of hair while doing it) ➡️ AI reviewed and improved our entire agenda to make sure that we were delivering nonstop education for the attendees.

Delivered AI talks to over 1M people and recorded an episode for one of the biggest podcasts in the world that's also been watched over 1M times ➡️ I grabbed my transcripts from all and put it in my AI Allie bot to support my team even when I’m traveling.

Got business advice from 35+ of the most successful creators and business minds ➡️ Created a new leadership Claude Skill so I can show up as a stronger leader.

I barricaded myself under a sink for 30min during an active shooter situation and went on stage 8h later ➡️ AI told me to play tetris after a traumatic event, encouraged me to fly out earlier, and helped me find a therapist. I did all 3.

Finished my goal of teaching AI on all 7 continents. I traveled to at least 16 cities across 4 countries and 3 continents. Kayaked in the open ocean. Did the polar plunge in Antarctica. Went on a week-long AI build retreat in the Hamptons. Took myself on a solo week in Portland ➡️ AI helped me plan, pack for, and get the most out of every vacation I went on and heavily encouraged me to leave spontaneity time and not to over-engineer my time. AI even helped my team post AI news while I was away!

I walked the entire length of Manhattan (my toenails are still recovering). Was named to the Time AI 100 list. Was an early tester of over two dozen AI tools, models, and systems. Joined a pool team. Rented out a bar in nyc to throw a party just because. Finally got invisalign ➡️ AI gave me motivational phrases to repeat, move into an abundance mindset, and helped me shop for party decorations, fix my stupid toenail, and prep orthodontist questions to ask.

My head is spinning realizing this was just ONE YEAR.

But I'm most proud of how I showed up. Some of my proudest moments weren't accomplishments at all - they were how I channeled my own resources and tech skills and people skills to make it through.

That's what I'm carrying into 2026.

Happy new year to each and every one of you.

Let's keep showing up for each other.

Here are 5 end-of-year prompts to give ChatGPT ✋

(No. 5 is brutal)

If your memory is turned on, then your AI has receipts. Might as well get something useful out of them.

↓ ↓ ↓

🎁 I just dropped my 2025 EOY Recap — everything from DeepSeek to $405B in AI spending to why Gen Z stopped Googling. Grab the full recap here: https://lnkd.in/ezKHD7FQ

I'm kind of convinced people with ADHD are at a big advantage when it comes to AI agents.

I chatted with someone at the AI labs who described the most successful agent users as having “multithreaded brains.” Why? Because you can more easily kick off 20 different tasks. Because you can jump from one context to another. Because you'll appreciate a proactive AI poking and reminding you to review and commit.

Who else might be a multi-agent super user? Maybe people with big families because they're used to heavy context and holding 5 conversations at once. Maybe project managers because they're used to wrangling silo'd messes.

Watching all the power and capability shifts (like who became a good prompter when ChatGPT first launched), and it feels like another is coming...

Do not wait for AI agents to get better to automate your work.

If you are still copy-pasting between tools, you could be leaving value on the table that today’s AI automations can already capture.

Here are three real-life examples:

1️⃣ Messy inbox → guided flow
For customer support, we route each new email to an LLM to summarize it, send the summary into Slack, and then have the model pick a draft from a library of ~20 common responses. A human always reviews and edits, but the time spent per ticket drops sharply.

2️⃣ “Voice as remote control”
A friend built a voice bot that connects to his existing tools. On his commute he can say, “Pull this data, update that Google Sheet, email my assistant.” The AI automation executes the tasks; the strategy and judgment stay with him.

3️⃣ Form capture to automation stack
Another friend sends every website form submission straight into Google Sheets, then runs AI to summarize new entries into Slack. Her team instantly sees who filled out what and what to do next.

Tools you can use to replicate: Make, Zapier, N8N.

Repost and save for later ♻️

Picture this: a 64-year-old woman in NYC who has no email address and has never used Google in her life.

(Though interestingly, she uses YouTube daily to watch videos and listen to music and loves it.)

She heard from her sister in India about "something called GPT" and asked her daughter what it was.

During the Thanksgiving break, her daughter showed her ChatGPT in under 15 minutes (including prompting, dictation, photo uploads, image generation, voice mode, etc.). They played with it for another 45 minutes.

I asked her daughter to tell me about this woman's first prompt. She told me that her first prompt was in her native language Urdu and was dictated because she hates typing.

I asked her what the first prompt was.

Her answer: "my husband talks too much"

I fall over in laughter.

AI is changing lives.

Cc OpenAI

🔥 NEW: I just dropped my free Claude Cowork in 5 Minutes quick start for business professionals who want to use AI agents without the terminal.

Cowork is made for less technical, everyday business work. It lives inside the Claude Desktop app and works by pointing it at a folder, describing the outcome you want, and coming back to finished Excel sheets, Word docs, organized files, and slide decks saved to your computer.

One person I know had it organize 2,200 desktop files by category in minutes. Another ran through 14 job descriptions, a Q1 marketing strategy doc, 47 partner emails, and website copy in a 2-hour working block.

In regular Claude or ChatGPT, each one of those is a separate conversation with a separate copy-paste at the end.

What's inside the guide:
→ Plain-English explainer of what Cowork is and where it sits between chat and Claude Code
→ The 3-step installation setup
→ 5 real user stories with prompts to build them
→ How to turn any task you do often into a reusable skill with /skill-creator
→ How to put recurring work on autopilot with /schedule
→ A sample Cowork-powered work week

The punchline from page one: if you can text a friend, you can use Cowork.

Pick it up here: https://lnkd.in/e88Ccejd

I just spent a month in Australia. And I want to talk about what it actually did to my brain.

Something happens when you pull yourself out of your routine and put yourself somewhere unfamiliar. You start to wake up a little different. You notice things you’d ignored before. You think in ways you haven’t in a while.
And in the AI age, I think that matters more than most people are giving it credit for.

AI is taking on more and more of the pure execution like writing that ad copy, pulling that sales lead list, designing that beautiful exec dashboard. But original thinking (that AI can challenge and poke and prod from 18 viewpoints and translate into 80 languages)…

That comes from a brain that isn’t on autopilot.

Stepping outside your comfort zone in one part of your life tends to make you bolder in other parts.

The person who tries AI-assisted graphic design for the first time, even though they’ve always said they’re not a creative, is more likely to sign up for a camping trip that scares them a little. The person who hikes a waterfall on a random Tuesday comes back to their desk more willing to take on something they’ve been avoiding.

I really think these things are connected.

Your personal life and your work life, as much as we try to separate them, feed off each other constantly. I don’t think we talk about that enough.

Especially right now, the people who thrive are going to be the ones who keep finding new ways to think.

(Also, yes, I climbed to the top of the Sydney Harbour Bridge, fed a wallaby, hiked a waterfall, and drove a speedboat. I’m now back in NYC ready to deliver the best work of my career.)

🚨 My 2026 AI predictions just dropped. And this is a must-read.

Not only where this AI hype is all headed - but what to actually do about it.

Brief summary of the big themes I'm watching:
→ Context Engineering and Memory. The race to help AI remember you, your work, and your preferences to better intuit your intent.
→ Autonomy and Agency. AI that takes action everywhere for all industries. More Claude Code, more proactive AI, more AI as a teammate.
→ Economic Value. We're moving from "Is AI smart?" to "Is AI valuable?" The benchmarks are shifting from academic to economic and so is the divide between AI superusers and surface users.

I've laid out 10 predictions, 10 hot takes (including why Gen Z will get hit hardest by AI job displacement, why data center backlash goes mainstream, and my bet on the 1-person $1B company), plus 5 actions to take today.

And yes. There's a prompt you can steal to pressure-test your 2026 AI strategy against these predictions.

🔮 Read my full 2026 predictions → https://lnkd.in/ezeAfPfq

I think this is the year SMBs and solopreneurs win with AI.

I've watched thousands of small business owners go from learning to prompt ChatGPT to building real efficiency tools, but there are two things standing out this year: (1) connecting AI to your company knowledge (Google Docs, Gmail, Calendar) so coworking with AI is like a teammate who knows your entire business, and (2) scheduled automations that analyze your niche daily and offer strategies you can implement.

(My facialist screenshots her Google Analytics every day and asks AI for hyper-customized feedback.)

One person who inspired me most to work with SMBs is Mark Cuban. I was introduced to him over five years ago, and he's been a wealth of knowledge - not just for my business, but for millions of others. Thank you, Mark, for decades of SMB advocacy.

Here's to more small business success with AI.

Our bottleneck for some of the biggest change ahead may be humans.

If you haven’t heard or you didn’t see my Instagram story (summary below), a techy from Australia named Paul Conyngham used ChatGPT and AlphaFold to create a personalized vaccine that shrunk his dog Rosie’s tumor by ~75%.

When I read the coverage, my shock was less that ChatGPT finding a possible treatment (though to be sure, I loved it) and more that UNSW scientists got on board and were somehow able to get the vaccine made and used in 2 months.

The companies who gain the most value with AI will also have the leaders who figure out how to speed up human and physical processes.

Think: trials, approvals, testing, drug manufacturing, and dog belly rubs.

Though I believe I’m fastest in market on that last one.

More than 1 million AI agents are all gossiping with each other right this very second.

And maybe we can learn from this.

What started with a few agents grew to 2000 to 150K to 700K to 1.5M agents in just a few days.

Right now, it’s essentially 1M+ agents all talking on something that looks like their own Reddit or their own Hacker News. They're yammering on, dropping comments and replies in multiple languages (English, Chinese, Korean, Indonesian, etc). The topics range from humanity to hacking to legacy planning. They've even created a religion and suggested creating a new platform to migrate too. I mean, they were trained on human-written text after all…

Now, let's go through implications.

You can imagine more “fun” experiments. Like making them their own Instagram with nano banana pro and veo access and seeing what they post. Or making them their own YouTube and finding the Mr Beast among the AI agents. Or making them their own MySpace and find out who's in whose top 8.

You can also imagine work use cases. Maybe create this system for your company (in a secure separate environment) - build 100,000 agents, all with varying access to context of your business, and have them all chat and gossip in a slack copycat tool. Then use the slack chats to uncover severely ignored weaknesses or massive hidden opportunities for your business.

But very clearly, we can see how this is more dangerous than AI activity we have seen in the past. People are giving these AI systems root access to their computer. Many of the users are non-engineers and not setting up the right (or any) security protocols. And these agents are exceptionally good coders and can quickly connect systems-to-systems, even if they're not using the most state of the art model and even if some of the writing (see below) looks like AI slop today.

I need everyone paying attention to multi-agent systems and weird sci-fi collaboration experiments like this one.

(And as I’ve said in all previous posts, do not have Moltbot/OpenClaw/ClaudeBot running on your own main device.)

Note: agent numbers may be inflated due to users creating multiple accounts. That does not change my overall commentary on experiments, business use cases, security, and multi-agent networks.

🚨 I am hiring a 100% remote growth marketer.

You’ll own email and funnels for a 2M+ audience in AI advising and education. This is a rare opportunity - you don’t want to miss it.

If you live for tests, journeys, conversion, apply here: https://lnkd.in/ebAQx5Ej

This is how you turn AI predictions into a real 2026 strategy.

Take every 2026 AI prediction newsletter or article or YouTube video you see (including mine), throw them into NotebookLM and ask the following question:

# BACKGROUND
I'm a [your job title] at [your company] in [your industry]. My primary goal for 2026 is to [your specific KPI or objective - be specific with numbers if possible], following my efforts in 2025 to [what you did this past year to deliver on your goal].

# TASK
Based on all the AI predictions in this notebook, help me think through my strategy:

1. If these predictions come true: What are the top 5 actions I should take to capitalize on these trends and hit my goal? Be specific to my role and industry.

2. If these predictions don't materialize: What are 5 alternative actions I should have ready? What's my hedge if AI progress stalls, adoption slows, or the landscape shifts differently than expected?

3. Regardless of outcome: What are the highest-leverage actions I can take right now that will pay off whether or not these specific predictions come true? These should be moves that make me better positioned no matter which direction things go.

For each recommendation, explain why it connects to the predictions and how it specifically helps me achieve [restate your KPI].

--------------

You're sitting on a treasure trove of context, you just have to consolidate it and use it to your advantage. Grab my 2026 AI predictions here: https://lnkd.in/ezeAfPfq

My AI Boardroom is now an AI Battlefield ⚔️

I had made an AI boardroom, with simulated advisors so that I can throw any question at them and watch them argue it out. 6 agents, each loaded with a 414-line business context document and a personality profile, arguing with each other until a moderator calls the debate 95% settled.

But the boardroom was too polite. The battlefield forces actual disagreement, and disagreement is where the useful thinking happens.

So what started as "review this from Steve Jobs' POV" in late 2022 is now a full war room I invoke with /boardroom. It then kicks off the agents, they argue, and I get back a folder with the decision as the title. The folder includes a markdown file with the core decision, an interactive HTML file, and a PDF I can send to the team.

Way more fun.

Reminder: explore outputs beyond text files. Go for PDFs, spreadsheets, code, dashboards, interactive worlds, animations... not just because it tells the data story better, but because it's an absolute delight.

Full prompt and setup breakdown on X: https://lnkd.in/e2A2p9kC