Reading up on ai-agents

I spent 48 hours in SF hanging out with billionaires, data labs and AI founders and execs running $1B to $100B companies. My observations: 1. Sooo MANY stories of data brokers making like $2M a WEEK selling company data to labs. I can't tell if it's the beginning or a bubble, and I kept wondering which niches are left. Digging into it actively. 2. At dinner, someone asked the table, "If a new model drops next month, does your company get better or worse?" HALF the table went quiet! I think that's the best question to ask about any AI startup right now. 3. The bar people talk about now is hitting $1B in revenue in 12 months (I guess that's the new unicorn?). What I didn't hear much about is retention. AI revenue can come in fast and leave fast, and I think we're going to see a lot of stories of AI companies with huge revenue and terrible retention. 4. It's pretty clear we're living through the reinvention of the computer. The model is the CPU, the harness is the operating system, and skills and plugins are the apps. Thinking about it that way makes it easier to see where the whitespace is. 5. We're maybe 12 months from an agent that can watch you work for a week and then do your job without any instructions. The screen recording plus agent memory plus local model combination makes this possible right now. 6. Agent harnesses feel like the new GPT wrappers, in a good way, especially vertical ones. The best teams I talked to treat the model as swappable. One founder walked me through routing different models to different jobs: one for taste and visuals, a cheaper one for the everyday stuff, another for text. They use prompts they can move across models over fine-tuning, and A/B test the output on what users actually do with it: do they edit it, export it, share it? A vertical harness is basically four things: the connectors for that industry's data, evals written by people who know the work, rules for what the agent can and can't touch, and a clean screen for a human to review. The model is maybe 20% of it. 7. There's a lot of reverse engineering going on/cloning going on more than I've ever seen. People are decompiling software and rebuilding it with agents, and someone showed a suite of like 10 apps that are worth $5B+ each all that he cloned open source. Me and everyone are still thinking through what it means when software gets this easy to copy. 8. My guess is the value moves to data. If the software can be rebuilt quickly, what's worth paying for in a legacy SaaS company is the years of customer data and workflows inside it. Some SaaS companies are secretly data companies. Gets me thinking about how do you truly value data for this saas companies? Like if we were buying these dying saas companies at 2x ARR before (as example), maybe there is a better metric. Idk, just thinking out loud here. 9. It feels like a good chunk of legacy software won't make it to the agentic era, especially anything built around people clicking through a lot of screens. 10. Every company I talked to wants to pipe its internal data into one place and expose it to agents through MCP. Then they look at the data and realize it's a mess: five versions of the same customer, old Slack channels, PDFs nobody can search. Cleaning that up feels like a big consulting opportunity! 11. A lot of founders are building the basic building blocks for agents: APIs, data layers and tools. Some of the companies I saw barely have a UI because the main user is an agent. I really believe the startups that build primitives to this agentic era will do extremely well. Building blocks is the best way to think about it. 12. Labs can get plenty of training data, but it's harder to get real examples of agents failing inside companies. Whoever has those logs has something valuable. 13. Someone running a big consumer company told me the next fight is over memory. The model that remembers the most about you wins, and the smartest model matters less than people think. 14. Physical AI came up a lot. Astra I think was the first model that got people into AI can actually do CAD, and I think a lot of people are thinking what does vibe manufacturing really look like. The line I keep thinking about came from a robotics founder: the hardest problems in AI are deployment problems now, not research problems. Their robots work on their own, call a human when they get stuck, and every one of those calls becomes training data. 15. Cybersecurity companies are having a moment and are actually cool companies to build now, especially the ones that clean up after a breach. Every agent with access to your systems is a new door into your company, and a lot of those doors were left open. 16. Some startups are worth more dead than alive now. A shutdown company with 5 years of Slack, email and code can get paid for licensing all of it. I wouldn't be surprised if founders start factoring that into when they shut down. 17. People are starting to talk about monetizing the agent web. It feels really early, but when the most popular user of the internet is an agent, what startups should you be building? 18. I think agent insurance is coming. Once an agent makes an expensive mistake, insurers will want audits before they cover anyone. 19. Several people said the hire they want most is someone with taste who can write evals. The models are getting good at generating ideas, so the hard part is judging which ones are good. 20. New trend I observed: startup raises VC, uses venture dollars to become a creator, startup fails, founder falls back on being a full-time creator and incubates other products bootstrapped. 21. Getting SIGNIFICANT distribution is HARD right now. Founders are sponsoring buses, buying billboards like crazy and throwing parties with famous DJs to get noticed. Kinda feels dotcom era in that sense. Attention is what's scarce! The growth guy going viral sharing his UGC strategy from Wispr Flow is a good example. It's the sauce everyone wants, and I truly believe there are probably only 50 people on the planet who know how to scale these types of businesses efficiently in this era. Always good to get that dose of SF. I felt sooo lucky to have such an incredible trip during such an incredible time in building businesses. Surreal in so many ways. Happy to be home and locked in. Until next time!

@gregisenberg

Introducing Underdog x Stripe Shop with Underdog. It finds the best option, saving you money and time. In a local browser, it gets a one-time Link card for the exact amount, and waits for your approval The first on-device Agent Payment in Underdog: Your Private Personal AI

@UnderdogAI

It’s pretty clear to me that the internet is headed toward a new business model that we may not have figured out yet. Consumers will increasingly outsource tasks to agents, and agents will work around the clock to execute them on their behalf. These tasks require inference spend. The most obvious way to cover these costs is by passing them through to the user. Users pay a monthly subscription or simply pay for credits that get deducted as inference gets spent. But with the cost of inference dropping over time, this will ultimately be a race to the bottom. Pass-through pricing will get commoditized and won’t be able to sustain margins. So startups and incumbents will have no choice but to offer a different business model. Paying for task-based outcomes is an obvious iteration. You can imagine customers paying agents when they achieve certain goals, such as successfully booking an appointment or executing a refund the user wouldn’t have taken care of on their own. But most tasks aren't worth paying for one at a time. The most scaled business models on the internet let everyone participate, even when they can't pay directly. This is why advertising dominated as the primary business model of the past 30 years. Incumbents will subsidize usage dramatically over the next few years to onboard as many customers as possible. We’re already seeing this with Muse. But the question is, how will this work get paid for long-term? Can agents adopt a form of advertising that covers the cost? Or do we need a completely new model where the few big paid outcomes subsidize everyone else's usage? We may be headed toward a new world of haves and have nots. Or a world of abundance where everyone can get anything done at any time for free.

@mignano

Great read from Patrick, as always. “Personal agents will, I think, act as a kind of structural subsidy for product quality.” Yes, and well put.

@tobi

before I joined Anthropic, I spent about 2 weeks hacking on this as a side project with Opus 4. It used the Agent SDK so it needed a constantly running process & didnt work that well but one prompt to Opus 5.5 ported it to Claude Managed Agents & made it way more reliable

@trq212

Assorted things I've been mulling over regarding agents in the economy: • Companies are in some sense the original superintelligence. It's often the case that they can use their amortized cognitive surplus to advance their interests against those of certain consumers: relative to the optimization capacity of companies, consumers can be inattentive, underinformed, irrational, etc. (Think, say, of some of the classic insurance upsells at the end of a booking process, which is typically priced substantially above what a regular policy would cost.) Herbert Simon introduced the idea of bounded rationality, and firms often end up benefiting from it. Thanks to personal agents, we will probably get somewhat more rational in how we engage with companies. • How exactly will outcomes change? Well, certain kinds of price discrimination will work less well (how do coupons work if every agent hunts fastidiously for them?), and certain kinds of customer cross-subsidy will change (today, people who forget to cancel subscriptions subsidize those who don't). Taken across the economy as a whole, I'm not sure what the incidence of all of this will be. • In corporate strategy, people have long debated the relative importance of distribution and product quality. Perhaps you've made something better, but does anyone know about it? Won't people just continue to buy the ACME Corp Mousetraps? Personal agents will, I think, act as a kind of structural subsidy for product quality. How would markets look different if individuals always spent at least 10 hours researching their purchases? On the margin, focusing on making a product better is, I think, going to become a more effective strategy for companies. This is good! • It true, a corollary is that information about product quality will become more important and impactful. Today, most people have very limited ability to aggregate and integrate such data, but agents will have insatiable thirst for such signals. (It's often said that data is becoming more valuable because it helps in the training of models. That's obviously true, but I think there's a secondary way in which it's more valuable simply as a complement to cheaper intelligence.) • A big question is whether agents will increase superstar effects (rational agents all settle on the same product) or increase dispersion (because everyone has slightly different tastes and preferences, which agents are good at eliciting, finally overcoming the artificial clustering we've been subjected to). • Many internet companies have monetized demand routing, which is in turn often predicated on short attention spans. "I'll click on the first thing I see." This has led to some amount of inefficiency, as companies try to propitiate capricious ranking algorithms and incur various placement fees. It's not all deadweight loss (willingness to pay is itself a kind of signal, which is valuable to buyers), but it's not great. How will the nature and value of this routing change in a world of personal agents? I don't know the answer, but I think that the effects might be pretty big. • Much of what I'm writing hinges on the assumption that personal agents will be on the consumer's side. I suspect that that will be the winning strategy, but whether it's true, and what it means to be true, involves a lot of other questions. Taken as a whole, the market is a massively parallelized, decentralized, and highly imperfect value attribution machine, and, taking stock of all of the effects, it seems to me that the quality point may well be the biggest: that agents will change the nature of the backpropagation such that the rewards for better products increase. Maybe this is too optimistic, but I suspect (and hope) that companies that make things that they're proud of will feel that the universe is a little more partisan in their favor.

@patrickc

OpenAI’s Dots launched last week and they've taken over @every. A few things that stand out: They protect your attention. Over the weekend, my Dot posted my model-testing results to Slack and brought back replies. I could keep working without opening Slack and disappearing down a rabbit hole. My weekend stayed pretty clear and calm. They’re useful for the little things you miss. People at Every are using them to pull important information out of school emails, catch missed Slack messages, and flag account-access requests. Voice makes them useful away from your laptop. Our head of video was talking his Dot through video edits while running around getting our launch out. The ecosystem matters. Dots feel roughly at feature parity with Muse, Instinct, and Grok bots. But having an agent connected to the ecosystem you already use for work and life is a real advantage. If you use ChatGPT, I recommend trying one. Permissions are still frustrating. I have to approve the same basic thing multiple times, and sometimes even after I approve it, my Dot says it can’t do it. That needs to get fixed. I’m less convinced we’ll stay loyal to the characters. Mine is named Boo, and I’m attached to him. But I felt that way about my OpenClaw agent, too. When something better comes along, I switch. In a year, I expect to be using a persistent agent. Boo will probably be dead. RIP, Boo. I get into all of this at the start of this week’s podcast, and then sit down with @bigwilliestyle to talk about where we go from here. In a lot of ways, Dots and GrokBots remind me of a wave we went through with OpenClaw eight months ago. We talk about personal vs. company agents, why we built our own company agent, and whether companies will end up with a whole shadow org chart of agents. Full episode below.

@danshipper

Embarrassed to share this, but it scared the shit out of me Last Thursday, an AI agent posted my personal bank balances into our company Slack. As me It broke down all my expenses in detail. Then it apologized Be careful of the dark side of proactive agents More below...

@ShaneMac

Your side projects should be compounding very fast with AI. Every problem solved can be re-used or extracted into a generic module. Every design decision should make the next thing beautiful by default. Your agents should be constantly stealing from your past decisions.

@brian_lovin

OpenSource Email inboxes for AI agents You can self host easily using @Cloudflare Check it out > https://goshenemail.com > https://github.com/boringcomputers/goshen-email

@Rasmic

Just connected my ChatGPT Dot to iMessage using @thelinqapp. It's exactly like Instinct, except it actually has context on me + can use my computer + has my entire Gbrain knowledge graph and all my plugins. Instinct = feature not a product.

@awilkinson

My biggest takeaways from @thsottiaux: 1. Most actions on the internet will soon be taken by agents, and Tibo doesn’t think most people have priced this in yet. When Notion launched its MCP server, it saw a flood of traffic, which strained its systems and forced it to rethink its economics. Every product team will soon have to decide whether to build for agents. You can hold off for a while, Tibo says, “but it is inevitable.” 2. The model picker is likely going away. One of Tibo’s biggest frustrations with the current state of AI is that users have to think about which model to use and what reasoning effort to apply. “I myself even get fatigued with the model picker,” he said, and he wants to get rid of all of it as quickly as possible. Dots ships with no model picker at all. You just talk to it. 3. Build as if models will be 10 times better in a year. Tibo thinks models will get cheaper and faster “at rates that are quite incredible,” and that all modalities will finally work together seamlessly. When he looks at what people are building, his reaction is often “you’re not quite getting it.” If you imagined “all of this being roughly 10 times better than it is today,” what would you build differently? His own forecast was too slow: he expected today’s capabilities a year or two from now. 4. Loops, graphs, and fine-tuned prompts are not the long-term solution for getting the most out of AI. There was a period when people got very excited about configuring agent loops and building intricate graph architectures. Tibo thinks that era is ending. 5. While Dots got most of the attention at DevDay, Tibo thinks the plugin ecosystem will be the sleeper hit. OpenAI launched with 16 partners for “Sign in with ChatGPT,” and popular third-party plugins will share in revenue when subscribers spend tokens with them. ChatGPT recommends plugins inside conversations based on retention and quality, which can put a good one in front of “a significant slice” of its 1.2 billion users. Tibo’s advice to developers trying to get discovered: “Build a good plugin.” 6. Tibo’s Dot warned him that production was down five minutes before the DevDay live demo. The agent figured out that DevDay was happening that day, that the demo was minutes away, that the demo probably depended on that production system, and that Tibo would want to know. It pinged him just as the launch was about to begin, and it even offered to fix the outage. His reply: “I don’t think you’re there yet, little Dot.” 7. Tibo’s agent teams grow and shrink with every model release. When he’s pushing the frontier, he builds larger and larger teams of agents. Then the next model lands, “a bigger agent can just do all of it and keep everything in memory and learn,” and he shrinks the team again. Expansion, then shrinking, then expansion. Don’t get attached to your setup. 8. Skill trending down in value: typing fast. Skill trending up: taste. Tibo says the people thriving now deeply understand the user, learn quickly, and “know what good looks like.” That’s part of why OpenAI is full of former founders, more than 120 of them, which makes it feel like “a mega startup.” 9. Tibo changed his mind about new grads. He hadn’t anticipated how fast younger people would embrace all this change and figure out how to harness it, and that shifted OpenAI’s hiring strategy. His example is Ahmed Ibrahim, hired as a new grad, who built much of the Codex harness and is now responsible for OpenAI’s compute fleet. What set Ahmed apart: he’s “incredibly kind,” “incredibly collaborative,” works on what matters without putting himself first, and is “a sponge,” using the latest tools to learn at speeds Tibo hadn’t seen before. 10. We’ll likely address the burnout associated with AI. There’s pressure to run 30 agents at once and ship more because you can. Tibo wants AI to “reduce the noise” and let you put your attention where you choose. Every time he fully disconnects for a week of holiday, he starts thinking in more creative ways, and he wants to bring that into the workday. Maybe you need fewer meetings, and you’d be more productive if you were better rested. The industry still has to figure that out, he says, but that’s the promise of AI: “it’s not just one more prompt per second.”

@lennysan

Introducing gdp-ts: Ghosts of Departed Proofs for TypeScript. gdp-ts is a library, linter and AI skill for safer API design. Under this contract, sensitive functions require 'proofs' that the caller performed an authorization check. The typechecker verifies these proofs at compile time, preventing your team and agents from shipping catastrophic security (and other kinds of) bugs. While these patterns have existed for quite some time, especially in ecosystems like Haskell, ① human code review and ② cognitive and syntactic overhead made these solutions niche. The situation is now inverted. Agents are writing more code than we can review, and they *thrive* in tight loops with hard constraints that would frustrate us. We see this with the rise of Rust, borrow checker, code aesthetics debate and all. The README and examples model a real-world Vercel API product constraint: changing the password on a Project requires a proof of a certain role + a certain entitlement. Thanks to Matt Noonan and Ollie Charles for their research in this space.

@rauchg

"Not only am I uninterested in the company’s home device, I can, for the first time, envision a future where I don’t buy Apple by default. Indeed, this already happened..." Anyone into the age of agents will eventually realize Apple is a bad fit. https://stratechery.com/2026/apple-and-a-hackers-future/

@dhh

Muse, dots, Instinct, and the question every agent will be asked

@worldnetwork

Cloudflare have integrated Pi Durable in the agents SDK. Really happy how it turned out. Props to @mattzcarey et al. for getting this done on such short notice. It "just works" :) https://github.com/cloudflare/agents/blob/df9c0ef6c4541c5552b5ce4d6ba292162dbd0908/packages/agents/src/harness/pi/harness.ts

@badlogicgames

Figma doesn’t want you to use MCP Amazon doesn’t want you to shop with agents Reddit won’t let Claude access threads X won’t let ChatGPT read tweets It’s starting. This happened with APIs 8 years ago.

@wesbos

We’ve been improving Loops a lot to automate work that should be automated In @slack or @linear just tell the agent you want to make a loop for something. The loop can auto pick models, connect to external services through MCP, can be triggered or scheduled. Some example loops from our team: * Investigate incoming bugs and propose fixes with browser tests * Find new Sentry errors and open fix PRs * Turn meeting action items into issues * Monitor engineering alerts and unusual cost spikes * Surface untracked container vulnerabilities * Scan and remove old feature flags * Watch feedback on workflows or product areas and keep an running document, slack or other sources

@karrisaarinen

Who knew that 22 years ago when we launched Basecamp we'd have built the absolute best framework to work with humans and agents in the same place. Here's @bitsweat talking to Clawdito (an agent) in Basecamp, assigning it work, getting responses, asking it to follow up, creating PRs, the whole thing. And here's @jasonzimdars working with Marie (an agent) and JZ bot (another agent, his agent) in a comment thread on a message board. Can do this anywhere in Basecamp... on a to-do, card, file, document, photo, anywhere you can talk to humans async or real time, you can talk with your agents the same way. And they can respond the same way humans do, right in line, right where you'd expect, with all the context in the thread. We've been running and refining this in our account for a while now. Anyone can do this today using the CLI. But we're also making it easier for people to create agent-users within Basecamp. Stay tuned for a lot more on this front.

@jasonfried

Adapting for a world of software factories

@zachlloydtweets

Instinct is the most impressive products I have ever tried. Even more so coming from such a young and clear-minded founder. Why do I like it? It’s proactive, smart, whimsical and generally delightful to use and makes you really feel like you are interacting with a top-tier human personal assistant working 24/7 to ensure every part of your life goes smoothly. After having tried OpenClaw/Hermes and all the coding harnesses, none of them really compare. Instinct has implemented the absolute freaking best memory system out there. It remembers everything and brings it up in surprising ways even making connections between different memories unprompted. It’s also crazy how proactive it is and how it seems to bring up what’s relevant just at the right time. I have been using it for travel, shopping, booking things, coordinating with my gf and friends for events and playing tennis (different Instincts can talk to each other on behalf of the owners). It’s truly incredible.

@frankdilo

it's basically impossible for someone to just "show you their prompt" now, because everything is about references, skills and examples I often ask my agent to look at 3 other repos I've made first, search the web for references, use other AI APIs, etc.

@trq212

Meet the winners of The WebMCP Challenge. These 10 projects show what people and agents can build together when websites expose structured tools agents can use. 🧵 See the winning projects and the builders behind them:

@OpenAIDevs

Inspired by @tobi and his datatree I built the same thing but for AWS billing. It shows per resource breakdown and you can export JSON and ask your AI agent to review.

@brankopetric00

Your agent can handle significantly more than you are currently asking of it

@benjitaylor

i love prompting for a full walkthrough of a flow i.e here i had it show me all of the steps involved in oauth and found redundant steps, layout shift, and some weird ui state agents don't have to produce slop, they can also help you make great software

@RhysSullivan

Eduardo Sasso (@eduardosasso) / X

@eduardosasso

Bot detection & human verification will be one of the most urgent demands for businesses over the coming years. Agent swarms will suffocate every website and form; small companies and government websites are most vulnerable. There is a huge gap in the market for this right now. When we looked at what offerings were in the market to use at X, there was not a single company that brought together all the latest technologies so we had to do it all in-house.

@nikitabier

all of my imessages & email now run through a single agent interface. i still text, but i deleted every email client from my phone & canceled every email subscription or app i had. gmail is basically just a database to me now. increasingly so are the rest of google’s services. i.e. i never want to manually write an email, create a doc, build slides, organize a calendar, or manage files again. i just want to express intent & have the underlying systems of record updated correctly. i think now you’ll see apps disappear quickly. many will become databases at best or entirely obsolete at worst. the agent will be the default interface to your personal & professional life.

@signulll

Building a (Restrained) Software Factory

@sawyerhood