👋 Howdy to the 5,963 new legends who joined since our last edition! You are now part of a 404,585 strong tribe outperforming the competition together.

LATEST POSTS 📚

If you’re new, not yet a subscriber, or just plain missed it, here are some of our recent editions.

🪐 What Changes After Your Exit (Nothing?). An interview with Ronan Berder, exited-Founder & CEO at Wiredcraft.
💆🏻 The Ultimate Founder Mental Health Stack. Psychedelics, depression drugs, dog walks, training, meditation, Whoop and loads more.
🔥 The CFO Who Killed The Spreadsheet. An interview with John Glasgow, CEO & CFO at Campfire.

PARTNERS 💫

Most AI agents talk. Thesys agents show.

Build conversational analytics agents that respond with charts, tables, slides, and reports - not paragraphs.

No workflows, no code. Just interactive UI your users actually want to engage with.

Lightfield is an AI-native CRM that captures every email, call, and meeting automatically. Ask anything about your business. 

"Why are we losing to [competitor]?"

 "Which deals went quiet?" 

"Prep me for my call with Acme."

 Real answers cited from real conversations — not stale fields nobody updated. 2,500+ startups already on the platform.

Try Lightfield free for 3 months with code: OPENSOURCE53

_{Interested in sponsoring these emails? See our partnership options here.}

HOUSEKEEPING 📨

I don’t know what it is about Attio’s marketing; they just seem to get it right, all the time. Today, they released their GTM Atlas, a free resource built with some of the best operators in the industry, covering the full customer journey from lead capture to expansion. It’s valuable, but also just very slick and well put together.

For those who don’t know, Attio is the CRM I use for this newsletter. And it cracks. They are well ahead of legacy players when it comes to the AI-led GTM of the future. Anyway, enjoy today’s piece!

DEEP DIVE 🕵🏻

What Is Perplexity Computer?

When assessing which AI to use in your personal and professional life, things can be daunting. Do I use ChatGPT, Gemini, or Claude? If Claude, should I use Chat, Cowork, Code? At the end of the day, what we all really want is to do more, better, and faster. The good news is there is one solution out there where you can, in fact, have your cake and eat it too: Perplexity Computer unifies every current AI capability into one single system.

In layman’s terms, you give Computer a job, and it pulls in its most talented friends to suit at every stage of the execution. This way, you don’t need to worry that ChatGPT sucks at coding, or that Claude doesn’t design super well (until recently), because you are using them all at once. It’s kinda like calling on the Avengers of AI.

Luckily for you, the team at Perplexity has asked me to act as your guinea pig, trialing Computer, to see what we can create with it. But not before we give an overview of the AI and LLM landscape at large, accompanied by a handful of memes, and a smattering of ill-fitted cultural references and analogies. Let’s do this!

The backstory on Perplexity

I don’t know about you, but Perplexity seems to me like it’s been around for longer than three and a half years. Well, it hasn’t. It was founded by four founders with one bet in August of 2022. Aravind Srinivas (ex-OpenAI, ex-DeepMind), Denis Yarats (ex-Facebook AI), Andy Konwinski (ex-Databricks co-founder), and Johnny Ho (ex-Quora), launched the original product: a conversational search engine that cited its sources, to a reaction from the Internet that read: ‘Is this a Google killer?’

If the brainchild of OpenAI were out there making up case law, Perplexity would cite references. It gave you confidence in the answer. It cited truth.

_{*Note: I am well aware that the above quote was totally out of context; I just like it a lot.}

Because of their success, investors across the ecosystem were clamoring to get a piece of the Silicon Valley upstarts. Bezos and NVIDIA were in early. IVP, NEA, Daniel Gross, Karpathy, and Naval, all on the cap table at various points. Take a look at how they’ve progressed over time.

Total raised: ~$1.72B across 11 rounds, backed by the who’s who as they’ve moved from web app to mobile to Mac to Comet (their browser) to enterprise tier to Computer. Each release widened the surface area, all the way to today, and their release of Computer, the natural endpoint of every product decision Perplexity has made.

AI’s fragmentation tax

Most companies today are scrambling between 15-20 shiny new AI tools to stay apace with their competitors. They are using ChatGPT Plus, Claude, Cursor, Midjourney, or Nano Banana; a Veo or Sora subscription; Notion AI; a niche research tool; plus whatever vertical thing they pay for in their industry. Hundreds or/of thousands of dollars per month, ten tabs open, and a context-paste reflex that feels like 2009 muscle memory. This is called a fragmentation tax. And it’s a tax we are all paying daily.

In the world of AI, every tool is looking to be your home base, but none of them are. Claude Code, and their more recent release of Cowork, are the closest thing in my life to a home base, and I can’t rely on it today because no matter how much I try, I burn through my tokens like Joey Chestnut burns through hot dogs.

There are two strategic camps today: 1) First, is the model camp, primarily OpenAI and Anthropic believing the path is owning the smartest model and pulling everything into it, and 2) second, is the orchestration camp, Perplexity (and a few others that for the sake of today’s post, shall remain nameless) choosing the path of being model-agnostic and routing every job to whichever model is best at it. I wrote in The World’s Hottest Industry: Post-Training that models are converging, with the alpha being in how well you can train your model to differentiate.

Which is why Perplexity's bet here is, in my opinion, very smart. The model camp has to win every round of an arms race that grows harder with every quarter. The orchestration camp wins by not playing the game at all, and actually gets stronger every time a new frontier model ships. More models converging just means more great options on the bench.

Computer

The image below tells you most of what you need to know; Computer is the conductor standing in the middle of the orchestra, baton in hand, deciding which section to bring in for which moment. I don’t actually like The Avengers, but if you prefer to keep our running metaphor alive, it's the seat Nick Fury sits in with the call sheet, deciding which Avenger gets paged for which mission. Same idea, different uniforms.

In practice, Computer is Perplexity’s agent platform: you describe the outcome once, and it plans the steps, chooses models and tools, and runs the whole workflow for you in the background.

There are four moves Computer makes that, when you stack them together, are quietly revolutionary.

• It plans the job for you: You write one prompt, like ‘Build me a competitor pricing dashboard for the top 10 AI companies’ (yes, I see the irony). Computer reads that and works out the steps before it lifts a finger. It decides: research first, run the numbers, design the dashboard, ship it as a working web app. Most users will never see this part. And that's deliberate.

• It picks the right model for each piece: Coding goes to whatever's best at coding right now. Deep research goes to whatever's best in research. Images go to Nano Banana. Wide search goes wherever wide search is sharpest. Nineteen frontier models on the roster.

=

• It's on no one's team: This is the bet I want to underline. Most AI companies are racing to be the strongest hero. Perplexity is racing to be the one holding the clipboard. If Claude 5 leapfrogs everyone on reasoning next month, Computer routes more reasoning jobs to Claude 5. If Gemini 3 ships a video model that cooks, Computer sends the video work there. You, the user, are never on the wrong side of the leaderboard, because Computer doesn't have a side. This is very important.

• It works while you don't: Computer runs in the cloud, not your laptop (yeah, suck it Cowork). You write the prompt, close the lid, go for a walk, take a meeting, do whatever it is you do. When you come back, the work is built and waiting: Code, research reports, Monte Carlo simulations, slide decks, PDFs, backed-up dates on your dating apps (joke, sorry). All of it baked in the background.

Stack those four together, and you don't really have a tool anymore. You have something closer to a junior team that doesn't sleep, doesn't bill hours, doesn't argue, and gets better every time a new frontier model ships. Which all sounds great in theory. Now let me show you what came out the other end.

A quick note

This ‘one model to rule them all’ versus the Harlem Globetrotters of AI tools is a pain I feel on a daily basis. My company, Athyna, has tried to lead from the front with our AI strategy. Late last year, we hired our first AI lead, built out an AI council with one member from each department, and began running weekly watch parties and learning sessions for the team.

One of the most important decisions was which LLM to bet on for our enterprise account, which, for us, was OpenAI. The biggest name, the hottest narrative—until later last year—and a user interface + experience that we felt was more user-friendly for the team. Augusto, AI lead, would build with n8n and other tools for the team at large; it would be OpenAI.

I was happy, okay with the decision, up until Dario and Anthropic went on an ungodly run of feature-mogging Sam and OpenAI so hard they needed to buy a podcast. The recent release of Claude Cowork was the point that I simply couldn’t take it anymore, the writing was on the wall, we should be chips in with Anthropic. After some back-and-forth, we decided to shift the entire organization to Claude. It made sense, the team was excited, and we even had the blessing of my favorite AI educator, Ruben Hassid.

All these movements had me thinking, ‘Maybe Mythos can come and save us?’ It goes without saying that this is extremely annoying as a founder. And makes you wish there were a simple way to harness what you need when you need it. Enter: Computer.

What I built

The fun thing about playing around with the Computer for this piece was that the team gave me the direction to “go out and build something,” rather than just play around with the tool, so that’s what I did. I thought to myself: What is a web app that is fun, useful, or interesting that I could build? A lot of ideas ran through my head.

Build #1

The first that came to mind was a Michael Jordan versus LeBron James career-comparison app. I wanted an app that showed the key moments and claims in the careers of both, along with straw-man + steel-man arguments for each. The next idea was a website that pulled the most popular fan fiction stories from A Song of Ice & Fire (Game of Thrones) lore and presented them to me in chronological order. The final idea was something I actually needed: a guide to Blackbird Sunrise, the event I attended in Sydney this week.

The event was hosted by Blackbird, Australia’s premier venture fund (of which I am a proud investor), and while it was only one day, there were close to 30 separate events throughout the week, all held at different times across the city. There were also four main event stages with different speakers running all day, so I decided to build my own, custom-built event site to navigate. After a few back-and-forth questions, Computer went to work.

_{*Note: I found it super cute that Perplexity was giving Blackbird’s design a massive pat on the back while it was researching.}

What came out was a beautiful website that matched the energy of the event. Along with the list of the 28 activations—nine on-site sponsor lounges and brand bars (Stripe, Anthropic, AWS, Airwallex, Deel, HSBC, Vercel × OpenAI) plus all verified off-site community side events, a live Sydney five-day forecast leading into the main day, a time-stamped programs section, a must-see list, and my personal favorite, and maybe most interesting feature, a custom Sydney map with a save-to-itinerary function.

It was actually all one man could ever dream of for a micro-event site. But after looking hard at the design, I knew it needed one more tweak to make it really sing: It needed to be weirder. So, after a few back-and-forth prompts, I made what I called the Sunrise 2026: Side Events & Brand Activation Guide, AI Slop Edition. It now does the exact same thing but in the AI-sloppiest fashion.

In the time it takes you to order a coffee, Perplexity Computer built me the ultimate, event-focused microsite. Something that I actually used to navigate the event and its surroundings.

Build #2

The second thing I set out to build was a live Dream 100 Command Center for our Athyna Intelligence product. Intelligence, a new post-training focused arm of the business, was launched a few months ago, and in doing so, we also launched our Dream 100, the 100 (or so) accounts we are looking to land, along with a handful of contacts from each.

And with a few prompts back and forth to Computer, and a little bit of context, that is exactly what I got. The Dream 100 Command Center, or Athyna Intelligence relationship OS. The dashboard I built ranks the accounts we need to prioritize, provides the full list of POCs for each, and suggests which actions to take next.

Using Computer’s ‘connectors,’ I was able to pull data from multiple CRMs, Google Sheets, Notion pages, and Slack channels into a single unified dashboard. My next step, which I am admittedly still working on, is to operationalize the dashboard. This would effectively be me building recurring tasks in Computer to run our outbound efforts to book meetings, and/or set up guest interviews for Open Source CEO, our secret backdoor into these relationships.

This is all possible and much more. Computer, in this case, acts a little like the all-seeing, ever-present Three Eyed Raven from Game of Thrones, or if we want to stick with the Marvel analogies from earlier, J.A.R.V.I.S., right hand of another famous Stark, Tony Stark, of Iron Man fame. Really quite a beautiful thing.

What you could build + do

Although building out my Sunrise and Intelligence sites was cool, the cooler thing is what you could do with Perplexity Computer with a little more time. The real value is in moving a load of your agentic workflows across from a standalone tool: ChatGPT or Claude, to Perplexity. This would singlehandedly dismantle the ‘Oh no, the new model that we are no longer paying for’ problem discussed earlier.

The next real stage for me would be to integrate every tool I use with Computer (already did most, as they are one-click integrations), and then start working through my scheduled tasks inside of Claude, giving each one of them to the gods of orchestration.

The two problems Perplexity Computer solves for—model agnosticism and working in the cloud—are for real. Model agnostic sounds like a marketing phrase, but structurally, it might be the whole game. Build your workflow around any one model, and you inherit their roadmap, their pricing, their downtime, and their bad days. Computer inherits none of those. When a new frontier model lands tomorrow that beats the field at reasoning, Computer routes to it the moment it ships. No migration. No rebuild. No meeting about whether to switch. The whole industry doing what it does becomes a free upgrade you didn't ask for.

The cloud-based, user-side reality is the bit that matters. While Computer is working, you don't have to be. You kick off three jobs in parallel, close your laptop, take a meeting, and come back to find them all waiting for you. The first time it happens, you laugh. By the third time, it's the new normal, and you start wondering what else you've been doing synchronously that didn't need to be.

Both lines compound. Every quarter, more models join the field. Every quarter, more work shifts from synchronous to async. Computer gets more useful as both lines move, which is the kind of design choice you only appreciate in hindsight.

Future

A couple of years ago, the AI deliverable was a paragraph of text or an image generation of an ayahuasca-induced dystopian fever dream. Today it’s anything from a working web app, financial model, or Word doc with redlines, to a 50-slide deck, research report with cited sources, or dashboard of your most ICP-fitting LinkedIn replies.

The future, as always, remains uncertain. The foundation models will continue to build behemoth-like capabilities under their ever-growing umbrellas, while Perplexity and others will lead the ‘orchestra class’ of companies, picking the right horse for the right course. There is one thing I do know: We are in for an exciting agentic future.

Fun facts

• They made an unsolicited bid for Chrome: In August 2025, Perplexity offered Google $34.5 billion for the Chrome browser, which was nearly double Perplexity's own valuation at the time. Chutzpah, apparently, is a renewable resource.

• The name is a literal AI joke: ‘Perplexity’ is the metric machine learning engineers use to measure how well a model predicts the next token. Lower number, smarter model. Aravind named the company after the scoreboard.

• They also tried to buy TikTok: In January 2025, Perplexity submitted a merger proposal for TikTok USA, later revised to give the US government a 50% stake at a $300 billion implied IPO value. When Aravind says "we move fast," he means it across business development too.

• Annualized revenue grew 6x in 18 months: ARR went from roughly $80 million in late 2024 to a $500 million run rate by April 2026. Very few companies in any era have hit that curve.

Extra reading

• What is Vibe Marketing? The Complete Guide - October, 2025

• If You Ain’t AI First, You’re Last - November, 2025

• The World’s Hottest Industry: Post-Training - March, 2026

And that's it! You can follow me on Twitter and LinkedIn, and also don’t forget to check out Perplexity Computer.

BRAIN FOOD 🧠 

TWEETS OF THE WEEK 🐣 

TOOLS WE RECOMMEND 🛠️

Every week, we highlight tools we like and those we actually use inside our business and give them an honest review. Today, we are highlighting Granola*—an AI-powered notepad that takes, summarizes, and organizes meeting notes without using intrusive recording bots.

See the full set of tools we use inside of Athyna & Open Source CEO here.

HOW I CAN HELP 🥳

P.S. Want to work together?

1. Hiring global talent: If you’re hiring tech, business or ops talent and want to do it 80% less, check out my startup, Athyna. 🌏

2. See my tech stack: Find our suite of tools & resources for both this newsletter and Athyna here. 🧰 

3. Reach an audience of tech leaders: Advertise with us if you want to get in front of founders, investors and leaders in tech. 👀