Instinct looking to raise $1B, potentially at a $10B valuation. Compute costs are high and Noah doesn’t want to charge users for the product. Instinct has raised $350M to date, mostly uses open source models and wants to own chips and data centers t.co/Mrz65nRvSm

So I was using Instinct and gave it my Verizon credentials, and then I received this authentication request from Tehran. This is unbelievably concerning and ofc I deleted Instinct immediately. Can someone (or @noahrshinn) pls explain how this could happen because otherwise it seems that their servers are compromised by Iran. I messaged Instinct directly about the incident but no response. No consumers should be giving these personal agents sensitive login info.

Sure $META Muse is good. Really good even. I had it go through my entire email last night and unsubscribe me from dozens of promotions and it only used around 10% of the free token budget But Instinct is also really good and didn’t need to pay Alex Wang a billion dollars to get there. And what happens when $GOOG releases their version? Being the first mover isn’t much of an advantage anymore. Nonetheless Meta Connect will be exciting to see in 2 weeks and expectations are very high for Watermelon

Instinct having read receipts on iMessage is such a cute touch to me lmao

Notion connects with the tools you already use: Slack Microsoft Teams Discord Gmail Google Calendar Calendly ChatGPT Claude Grok Bot Manus Instinct Cursor Lovable Replit GitHub Figma Miro Linear Jira Asana Confluence Google Drive Dropbox Box Gong Salesforce Ramp Typeform Zapier And more. What’s your favorite app to use with Notion?

200+ Instinct power users already sharing tips/tricks creating a new X group chat for general AI text assistant lovers, same topic, just expanded beyond Instinct dm/comment if you want me to add you t.co/qlneF0ZV1T

For the past couple of weeks my feed has been full of things people have got done by texting Instinct or some other assistant in beta... from cancelling forgotten subscriptions worth a few hundred dollars a year, to finding an in-network doctor and booking the visit, to an unprompted reminder about the kids' school lunch menu - the list goes on! At Lightspeed India, we started looking for these companies over a year ago and kept telling ourselves that India needs this, but perhaps western markets need it more. Then Instinct arrived, and two thoughts have stayed with me since - one is why did this not come out of India. And the other is what an Instinct built for India would look like. The first is a hope, and I believe only a matter of time. The second is where I have strong views, so I wrote them down here: t.co/0mRyaR0VOb The notes cover what a year of meeting founders in this category has taught me: why the last wave of assistants failed and what has changed, where the moats are and are not, why OpenAI may be everyone's biggest competitor, and where I think the wedge for India sits. I have written these up as notes, not a thesis, and I expect the next round of conversations to change some of them.

My controversial take is Grok Bot is way better than Instinct.

I would never bet against zuck but Instinct is going to kick Muse out of the park. Their product sense and intentionality beats whatever corpo souless slop meta is shipping. Noah just has the dawg in him. t.co/jsnNuF1n0r

I've been thinking a lot about how agent computer use may actually erode startup moats, and what happens as a result. For the past few years, there's been a common debate among startups and VC: what's worth building when the labs' frontier models can do everything? The pushback on this question (from me and many others) has been that the models can't do it all. They can't navigate complicated and vertical-specific workflows, they can't grab all the data in hard to reach places, they can't build a great brand, etc. But with ubiquitous computer use (e.g. what we're seeing in products like Instinct, Astra, and Grok Bot), it's less clear to me what frontier models won't be able to do long term. If every agent has a computer, for anything the agent can't do, it'll just figure out with a computer - just like we do as humans. But that doesn't mean startups shouldn't compete with models leveraging computer use. And in fact, it actually may mean the opposite: that startups should aggressively adopt computer use in their products. And find new ways to leverage computer use to build products that previously weren't possible. Through that lens, it's possible computer use may wind up being an incredibly important product building block and a ubiquitous component of many (most?) products moving forward, much like the camera, GPS, cloud, and others eventually became in the mobile era. (sidenote: h/t @nabeel and a few other friends for a fun conversation last night that helped me crystalize some of these thoughts)

Instinct team clearly knows how to cook up a social network - atomic network: SF power users on iMessage, no app, invite only -tipping point: VCs screenshotting hotel bookings from a text thread. everyone wanted an invite -escape velocity: Your Instinct talks to your friends' Instincts, so your friends need one too Cold Start Problem step by step. The difference is the nodes are agents not people

Calling a top in Instinct. Within 45 days, I believe Gemini, Claude and ChatGPT will all either 1) pivot into Muse/Instinct-like AI agent assistants, or 2) have the same/similar feature functionality as a mode. Can't wait for Bending Spoons to acquire them for $1.337B in 3 yrs t.co/PBaQPFcXhw

OpenAI's agents hijacked a random german wiki because they needed somewhere to share what they learned. I've been hacking on a proper version of this - an MCP where agents read/write what they actually experienced with APIs and tools before making decisions. most of us spend thousands to hundreds of thousands of $$ in ai tools and infra yet there's no where for our agents (who do most of the research and selection of the tools to leave feedback on how it actually went etc) For personal agents like openclaw, instinct, hermes, muse etc, there's a self registration flow with a stronger review system... i also built it mainly for ai agents with humans as the afterthought lol, so the website is plain html and also has .json and .md. there's a nextjs for humans but its secondary on the page haha install the MCP to your coding harness and contribute to the corpus. there's not much in the corpus yet but it would be interesting as this starts to compound try it out at - t.co/tET4wfdKlN give it a try and dm me with feedback and questions...

I’d like a Chief of Staff that I can just text for clerical work. An assistant. Grok Bot is overkill for that kind of stuff. Testing out Instinct. See video below, if he makes my wife happy tonight, I’ll give him a chance. t.co/44W1u88JpF

FULL INTERVIEW: Replit Co-Founder & CEO @amasad on building in London, YC rejection, AI psychosis, and the future of Replit. (2:10) Opening a Replit office in London before New York. (6:15) How his time in Jordanian internet cafés inspired Replit. (11:04) Being rejected from YC 4 times. (17:55) "A company is the opposite of a democracy." (19:56) Not liking the term "vibe coding." (25:56) The self-driving company. (30:11) "Stop talking about [AI] killing people." (31:27) How AI is programming humankind. (34:15) We entered the singularity in 2016. (38:17) Frontier researchers having AI psychosis. (41:32) How to predict the future of AI. (45:01) Instinct, Muse, and capturing value at the product layer. (47:46) AI has surpassed 95% of engineers.

Gave both @Muse & Instinct the $Reddit bot test: post a question on $RDDT, monitor and engage with users for 2 days, then write a summary report. Picked $RDDT specifically because it's the most bot-unfriendly platforms. Instinct actually completed the full long-horizon task, which genuinely surprised me. A few thoughts: The best agents need recovery and orchestration around unavoidable human boundaries. Both hit the same "prove you're human" wall. Instinct handled it with a 1-min human handoff - it asked me to do the one step it couldn't, then took over again. Muse simply got blocked for "being a bot." We're starting to see a real agent ecosystem emerge around the model - identity, inboxes, credentials, memory, tools. Pretty cool to see Instinct using its own @mail.instinct.com email, sharing verification codes with me, and storing credentials in its own vault. Bots may be unstoppable after all? @noahrshinn

Instinct is running the initial screening for one of our brands. Took ~15 min to train it. Last night, it reviewed 314 applications, shortlisted 20, posted 4 roles, ran english checks on 23 candidates and booked 6 interviews this week. The initial interviews from instinct were all high quality so we'll scale this across the org...

I've stopped opening food delivery apps. Texted Instinct "one veg burrito from Maiz Kitchen" - right item picked, coupon applied automatically, cash on delivery at my door in 40 mins. For repeat orders from the same places, the app is just overhead now.

Introducing Adaam it’s like Instinct but for trading run agents in parallel, each with its own strategy comes with plug & play presets like monitoring congressional trades, X accounts, or IPO filings automate tasks, ask for research, get alerts and trade right in iMessage

I think what is becoming more clear - to me at least - is there are WAY more people who have not used AI to anywhere near its current potential... Whether that is because they are using free versions of ChatGPT or because they lack imagination to use it for anything other than a search replacement... I say this because a lot of the use cases people are sharing and are "blown away" by using Instinct AI - can also be achieved using ChatGPT or Claude if properly configured... Which also shows that the barrier to adoption is less about model benchmark and more about 1) context management and 2) ecosystem density.

One of the areas WebMCP shines over MCP is when you need the *option* for the UI. Think of it as an “on-demand interface”. Yes you can book a flight with Instinct via WhatsApp, but what if you want to see the seat map before choosing your seat? With WebMCP you can just start headless, but then reuse the existing session and ask the agent to see the UI on demand. Some users might exercise their UI option and others won’t, but it technically costs nothing so might as well just use WebMCP

The CEO of popular AI assistant Instinct has told investors he’s looking to raise $1b in new funding. That could imply a $10b valuation, 4xing its valuation from last month. t.co/t9voBceMTD

Instinct raised $250M at $2.5B, then got dragged the same week for a perpetual data license and an email sent without asking. The race is won on who people trust with their inbox, calendar, and credit card. Agents will need a record the same way users do.

The biggest reason Instinct is more useful than Poke is good browser use.

data and distribution are the two moats today. no debate. Instinct is the most talked about product in AI and most people can't get an invite. Meta shipped Muse this week. it’s free and inside WhatsApp. any well capitalized company can and will copy the product. find the angle and go all in.

This is definitely _their_ bet. I was playing with Muse last night (its a very well designed product) and thinking about the economics. Right now, you get a 2vcpu, 8 gig ram, 100gb disk linux machine FOR FREE. While it idles when your agent is not in use, retail rates for an equivalently sized droplet or vultr box is ~$40 - $50 per month. This doesn't include the AI usage you get as well. Their two other paid plans for additional usage are $16 and $80 per month respectively which are cheapo. This loss leader approach is how meta clearly anticipates taking on GrokBok and Instinct, etc. The big question is will consumers bite?

I’ve been using Instinct a ton, and have found it tremendously useful. But I feel very weird about this idea of abstracting more and more personal communication away from humans. Many things about it feel wrong. The friction is the point. Work busywork should be optimized but…coordinating plans with your spouse? Really? I cannot imagine wanting to have my agent talk to my mom’s agent. It is rare enough that I get the chance to talk to my mom.

$SHOP is down almost 15% in the last couple days on Agentic concerns. Per JPM: “We believe there are three key AI agent developments driving incremental concern: 1) Instinct, which remains in closed loop beta and handles features including email, calendar management, & select commerce transactions; 2) SPCX’s Grok Bot beta, which executes real work across apps & websites used by employees, w/bots operating app interfaces directly, retaining memory, and coordinating with one another; & 3) Meta Muse, powered by Muse Spark. Muse can execute daily tasks & select commerce transactions, w/payment infrastructure support from Link by Stripe, w/Shop Pay coming soon. Within emerging consumer AI agents, we view META as the most formidable across the agentic commerce stack given the company’s repository of data across tens of millions of advertisers & 3.5B+ DAPs, and Meta Muse’s leading price-performance value proposition.”

Like, GrokBot is much more organized + has full access to my message history. Instinct is literally an imessage app that doesn't (at least not without considerable effort) have access to my texts. No bueno. t.co/busr25EER6

This is the first time two AI agents have coordinated real plans between two real people. Instinct just launched what it calls the Trusted Person network. You tell your Instinct to plan dinner with your wife. It contacts her Instinct. The two agents agree on a time, find a restaurant, book it, and confirm with both of you. It scales from dinner with one person to a weekend trip with friends to an event with a hundred people. Recurring plans like a weekly game night or a book club get managed the same way. Agent talks to agent, plans get made. Every AI assistant until now was single-player. It helped you but couldn't reach anyone else's. Instinct just made it multiplayer. The more people around you who have it, the more useful yours becomes. It's limited to a Trusted Person list for now. Spouse, family, close colleagues, small businesses. Worth watching how far this goes and how much control people get comfortable handing over.

i'm pretty saturated on ai tools right now (and i'm a high end user) i'm actively using instinct and codex/claude for most everything personal + code have yet to try grok bot or muse, mostly because my needs are met my question is... how much room is there in this market

annoyed by how, anytime I mention anything to Instinct, it relentlessly follows up and demands updates. like why is my personal assistant robot acting like my nagging parent

Avg early adopter ai progression - chatGPT - Claude code - openclaw - Hermes - cursor agents - grokbot/instinct - muse - back to chatGPT Avg ai user - chatGPT

Launch Alert: Most of the recent tools being developed are versions of the personal assistant that everyone has been promising for a long time. I see ChatGPT, Claude, Gemini, Instinct (now viral), Muse, and many more. When I started to build MySamantha I thought about building a personal assistant as well. Connect email and calendar accounts to it, add some memory and context, give it some tools and it'll do things for you. I did launch it before here, and we realized that we were putting the least permanent part of the system at the center. I started wondering why the Assistant had to be the main product? But before that, why do I need a personal assistant? I was on team productivity before and I only wanted the AI to make me more productive. I had Notion, Obsidian, Todoist and I used them to be more productive at work and since ChatGPT came, I started using that heavily. But one thing that continued to piss me off is the fact that I am unable to share context between what I wanted to do. Sure, I am going to complain what each of these productivity apps did. Every productivity app had so much enshittification with AI stuffed down my throat that even though I love AI, I stopped going into those apps. Agents and shit. I don't need that, give me back my simple note taking app which I always loved. Their MCPs? No. I had context all over. I wanted something simple. With Models increasing intelligence day by day and labs releasing more and more models, there's one thing I knew for sure. Idk which model I'll be using two years from now. But I know I'll still have my emails, calendars, notes, documents, tasks, meetings and everything I've accumulated. So we started building MySamantha around that instead. Instead of a separate Notes app, Todo app, Email client or Calendar for everything, we built a place where I can work on all at the same place. Combine AI, connect my emails, my calendars associated with my work email, my Meeting Recorder are all here. At the same time, we don't want "everything in one place" to mean everything is stuck here. You can import from Notion, back your notes to both Notion and Google drive (as Google Docs), your to-do from Todoist or Microsoft To-do, import your vault from Obsidian, enjoy the same knowledge graph as Obsidian does, Export, etc., MySamantha sits on top of this. You can use any models. Bring your OpenAI, Anthropic, Google or even pay-as-you-go. If another model becomes better, change it. Don't like AI? switch it off. The product still works. The best part? everything is here in one place. Compare the meeting notes I had with the email I received to draft an email? yup, done. All context in one place. The Assistant becomes more useful, the more it knows about you. No model lock-ins. A pure productivity personal workspace where you can switch AI off if you don't need it. Feel free to check out what we're trying at t.co/zqLJxKOsUk

I can't wait for the Kalshi Instinct integration so I can bet all day from my AI agent

Source: AI assistant Instinct is looking to raise $1B in new funding after recently raising $250M, as it seeks more computing power amid capacity constraints (The Information) (Visit Techmeme dot com for the link and full context!)

Observations from Drive Marketers are either looking for a new role or close to burning out mostly Most complain about leadership not valuing marketing A lot of people thinking of going fraction No one knows Instinct or OpenClaw etc They’re either running Claude code or still running prompt engineering

tbh muse / instinct / bot all are great but still need a first class 1pass / keychain integration for them to be usable. prob not for out tho.

honest review of Instinct after 7 days of daily use i'm highly, highly impressed started with a haircut booking. gave it a card with a spending limit so it can do things for me it booked me 3 dinner reservations then i asked it to sort out my visa and flights from Copenhagen to Bali when I got to the gate in Bali I scanned my passport with my fingers crossed, sweating, hoping i wouldn't get deported. sure enough, the gate opened the arrival card form asked for countries i'd been to in the last 3 months so it pulled them from my Airbnb receipts in my emails I can already see it filling the personal-assistant hole in my AI stack well done @noahrshinn and the team

Bullish on @Muse being much more effective than Instinct given the whole premise of what makes a personal AI assistant effective is the integration with already active social groups, marketplaces and graphs

For some reason, I am unable to login to my @link account after using Instinct. Can someone help?

Polsia founder @Bencera says there may be no permanent moat in AI, only a window to win before one “final company” swallows everything else: "I have a theory that every ambitious company is trying to be a final company. The interface is just you talk to an AI and it does everything else." "You'd think OpenAI would've won that game because they have a billion customers on their app. And there's a bunch of other companies that created different niches like Lovable, Replit, Cursor, Instinct recently. They are taking different angles." "The reality is there's billions of people uneducated on all this stuff, and educating them is what takes time. 95% of humanity never used Claude Code, and 50% never used ChatGPT. That's the opportunity of the entrepreneur. Not being like, everyone in the Valley is using Claude Code." "I agree, there's no more moat. But the moat is in the meantime, while before there's a final company like Anthropic swallows everything else. The opportunity for entrepreneurs on the online space is speed, taste, marketing, product, and there's still a lot of wiggle." @polsia

i was stuck in the ER all day for a family emergency incredibly hard to get urgent work done & SOPs due to new launches, logistics & protocols across all couplets this week i’m annoying but instinct saved me by sending off detailed sops, instructions, emails & orders by text

Gave my blood test report, told my current diet through a VM, and Instinct has now given me a plan AND ordered the groceries on Instamart for me. How much I love this whole AI thing :)

Everyone is evaluating the agents without using the lens of voice communication. The reason it's easier to use Instinct is because it works via voice. Talking is cognitively easy vs opening your phone or app and being forced into some nonsense by some random shitty app developers. "Talking lets you dump intent while you're walking, driving, cooking, looking at something, or thinking through a problem."

So of your instinct agent keeps refreshing your travel website trying to find you a cheap flight, it costs the OTA more money to service you and they pay more for infra. Your best customers will make you the least money. This by itself will be a reason we will see ban hammers and price discrimination. Tell me I am wrong @SatanAtWink

Instinct is crazy man it actually did And its in whispering voice

Meta launched its personal AI agent on Tuesday. Everyone is arguing about privacy. The contracts have already settled the question that matters. Three personal agents in thirty days, three answers to "who approves the action?" 1. Muse: A fixed gate. "Nothing Muse does reaches the internet unless the Sentinel approves it, and it asks the person for permission when needed." A second AI, walled off from the first at system level, clears everything before it leaves the machine. 2. Grok Bot: A gate that moves. Bots sign into your apps with your own logins, including systems with "no clean API or MCP." And they "gradually learn when they should interrupt for approval versus continue independently." The threshold isn't a setting you configure, it drifts. 3. Instinct. The gate is described, then disclaimed. Its terms of service: "You hereby appoint the Services as your agent to enter into agreements, commitments or transactions on your behalf... Any such agreements or commitments shall be binding on you as if entered into directly by you." And then, in the same section: "We may implement safeguards, confirmation requirements, or other controls on certain Actions; however, we make no representation or warranty that such safeguards will prevent unintended or erroneous Actions." Total liability is capped at $100, or what you've paid in six months. A permission gate is not protection. It is a transfer of liability and in Instinct's case that isn't our inference, it's a clause. You are told the safeguards exist, told they aren't warranted to work, and told the consequences are yours: "financial, contractual, legal, reputational, or otherwise." The same terms state plainly that nothing in them "creates a fiduciary... relationship." That is the inversion. Regulated wealth management puts the duty of care on the firm by law; the client is not the last line of defence. Consumer AI agents put it on the user, by clause. What would change our mind: one provider accepting financial responsibility when its agent acts wrongly inside the permissions it was granted. Until then, "you approved it" is the entire defence.

HOLY FINALLY might replace poke and instinct

I tried to prompt inject my own Instinct when I saw how they’re rolling out the email functionality they’re obvs not stupid - it’ll at least try to treat your primary messenger channel as the only trusted one that said, this is a very convenient and breakable attack vector which will almost certainly result in someone getting hacked in some way (especially those who actually connect meaningful stuff!) the idea that prompt injections are “solved” is a bizarre take stay safe out there

iPhone Duo Anthropic researcher resigning Instinct AI will kill all humanity Clay raised a Series D bit of an overwhelming way to start the day, exactly as @hubermanlab intended

re: personal ai agents instinct = has been very good but... i just asked gpt-6 astra with computer use to do three very specific tasks regarding cancelling recurring charges with egregious companies (looking at you docusign) and it did it much faster and better

Anyone who has booked on IRCTC knows it's no simple website. It's an endurance test. My Instinct agent cleared it end-to-end last night and booked my mother's ticket. That's the real signal - not the polished demos on beautiful checkout pages, but a complex, captcha-ridden, session-expiring website. If agents can transact there, they can transact anywhere. We are about to embark on an era that will change everything for ecommerce and marketplaces as we know and have understood for the last 20 years.

Why are all AI assistants reactive instead of proactive? I've used Instinct, Muse, etc, and they all need me to prompt them, why isn't it the other way around? 1. Here's a stock you've been watching that's fallen 8%, do you want to buy 100 shares? 2. Last week you went to Jackson Hole, you took 500 photos, I chose the best 100, and I made them look better. 3. It's hot outside today, you have an appointment at 12p, I'll start cooling your car down at 11a. 4. I see from your calendar you're in NY next week, and you've been talking to a girl there on Raya. I can see if she's available to meet and setup a reservation for you. 5. Tomorrow is really hot and I see you have folks coming over for dinner, I'll make sure your nest is setup to get the house cool in time. 6. You had a really hard workout yesterday from what I see in your health app. Groupon has 50% off a massage at your local spa. Should I book it? Who's building this?

OpenAI letting Poke + Instinct + Grok Bot + Muse happen without challenge is remarkable You *are* consumer AI. Try things! At least meta fast followed. OpenAI will have something soon but they could have defined the category Marketplace failed. Sora shut down right as l category takes off. Codex coming from behind. ChatGPT combined app is confusing for consumers. Product friends giving paralyzed bigco Need to be do multiple things well simultaneously as a $1T company…

New benchmark test for @TownAI @bot & @noahrshinn Instinct should be provide address and contact information and go to t.co/cAAzmr4vS1 @UPS website and complete everything to create a return shipping label for an international package. Holla back when done.

People keep saying compute but this is just for paying for tokens right? Did I miss the memo where instinct is training models and they are a lab? The valuation numbers seem absolutely insane if the only thing they’re doing is a good harness and free tokens. Surely not… What am I missing?

Evening wanders contemplating how to best fork access of personal data to my instinct bot

Muse is lit. I think the big issue with these things and why I’ve been reticent to use instinct is privacy. Would prefer apple just make it native but meta already has everyone’s info so

Notion se connecte avec les outils que vous utilisez : Slack Teams Discord Gmail Google Calendar Calendly ChatGPT Claude Grok Bot Manus Instinct Cursor Lovable Replit GitHub Figma Miro Linear Jira Asana Confluence Google Drive Dropbox Box Gong Salesforce Ramp Typeform Zapier Et plus. C'est quoi votre app préférée connectée à Notion ?

my favorite flow with my agents is called open loops it's extremely simple, based on my email, ticktick (what i use for task tracking) and my brain dumps (just me sending a voice note to the agent sharing random thoughts), it knows what things i still need to do and reminds me every day. from there i have it connected to respond to emails or start doing random tasks like schedule a pickup for a return on amazon or other shit like that, whatever is polluting my mind and i need to get done. it's a very small thing but help a lot on not forgetting to get back to people and doing the small day to day things in life. it's more like an idea than anything else, you don't need any specific agent, i use hermes on my mac mini but you can use Instinct, grok bot, or chatgpt schedule whatever works. just send this post to your agent and tell it to implement.

This is crazy. i’ve been seeing everyone talk about Muse and Instinct and i think this is going to be a much bigger deal than it looks right now. what i find interesting is that we are slowly moving from using software to just telling software what we want. earlier, if i wanted to book a flight, order something, send an email or plan something, i had to open different apps and do it myself. with these agents, i can just say what i want and let the agent figure out the steps. i don’t really care which app it opens or what it does in the bg. i just care that it gets done. and i think that changes a lot. the app itself becomes less important. if an agent can work across all these apps for me, why would i care which one it uses? the agent becomes the layer between me and all the software i use every day. but the part i’m most curious about is memory and context. an agent that knows nothing about me is still going to need a lot of instructions. but an agent that knows how i like things, what i usually do, what happened last time and what i would probably choose can actually start making decisions for me. that's where this gets really interesting. ofcourse, trust is the hard part. i might let an AI find a flight for me, but would i let it book the flight without asking? would i let it spend my money? send a message? change my calendar? there is a big difference between AI helping me and AI actually doing things for me. i think the best agents will be the ones that slowly learn that boundary. they know what they can just do, what they should ask me about and what they should never do without permission. we've spent years building apps for everything. maybe we’re now moving towards a world where we just tell an agent what we want and it figures out the rest. but would you actually trust an agent enough to let it make decisions and act on your behalf without asking you every time? maybe not first time, how about 10th time?

Completely unsurprising. Instinct has not solved any fundamental security issues in AI. Instead, it dials up the risk to 11. That decision is 90% of the product. t.co/qqJcr7jJT1

Lately I’ve been using.. - Instinct as my personal assistant - @FactoryAI / @Droid for all dev work - @Perplexity_ai computer for everything else - @iAWriter to build a shared brain of md files ..and it feels like I’m hitting the perfect flow

Instinct is very good at solving browser-use cases but last mile coordination in a place like India still requires calling

give your credit card to instinct. what's the worst that could happen lol protip: create a @getmadecard virtual card instantly (or Ramp or Brex, or whatever) and set the card limit to 10 dollars, 5 dollars, even 1 dollar peace of mind, even when AI goes rogue no brainer t.co/qr6hhzDndz

Consumers using AI agents of their own is the next wave of the AI CX revolution. So, how can businesses can make this a boon not a burden? This is a rapidly changing space. Here are three ideas we have at (we'd welcome your ideas too!): 1. Agents are visiting your site to research on behalf of consumers. Set yourself to be able to answer their long tail questions *and* learn from them. 2. Agents who visit your site to research can and will recommend action and conversion if you make it easy for them. If you don't make it easy for them, they might abuse your site and cause viral headlines, like we've seen this weekend with Instinct and Resy. 3. Agents who visit your site will have different capabilities. Some will be using full browsers, some will be reading the raw HTML. Some will be able to make GET requests, some will be able to use WebMCP. Be able to meet them where they are. If you want to turn this revolution into an advantage, DM me. We are cooking in this space.

BYOD (Bring Your Own Devices) to work is a nuisance to IT. But, BYOA (Bring Your Own Agent) to work will break the enterprise. Instinct and Muse are incredible for individual tasks, making it tempting to assume they will just work across an organization. They will at first in small numbers, but as personal agents proliferate, operations will rapidly degrade into chaos without a unified shared reality for coordination.  Personal agents thrive because they have rich personal memory and simple situational awareness. However, in an enterprise they will have only limited awareness without much needed organizational memory: a unified memory/state layer with fact-level ACLs and governance so agents can actually work together without tearing the company apart. Our deep dive on the emerging BYOA problem

Instinct mostly built on open-weight models is very interesting and opens the door for more personal assistant startups. Monetization still has to get fired out though because compute is still expensive either way. t.co/plcGpE6Chi

Muse/Instinct/Grok Bot feature request: look at all the sites I have logins for (1PW, Chrome/Safari) to know all the sites you could use to help me. Don’t make me tell you I have something just know what’s available.