For the past few days I've been using Instinct, the text message assistant that just raised at a $2.5 BILLION valuation. It's an exceptional product, and the most proactive personal assistant agent I've used. But it's doomed and their investors are going to lose all their money. It's a perfect example of feature not a product. The moment that OpenAI or Anthropic launches an integrated iMessage agent, it's game over. Why would I want a personal assistant that doesn't have access to the memory/knowledge graph I've standardized on? But who knows - maybe I'm wrong. Would love to hear the counter arguments. So curious how their investors got there.

Had many conversations this weekend re instinct / muse / consumer agents, key topics: - Commerce is the obvious opportunity but unclear how to handle supply constrained markets (hot restaurant tables) vs demand constrained (people that need to fly to San Francisco tomorrow); auction + reverse auction is intuitive but doesn’t capture loyalty / spend / LTV / brand - In a world where agents are the primary consumers what are their incentives? How do you market to them? What are their preferences and how distinct will they be from humans? - Will the first breakout multiplayer pattern be agent to agent (instincts trusted friends) or agent inside human chats ? - How do you design agents that increase social capital? - What are the patterns for comparative advantage / agents hiring other agents and what are consumer preferences in terms of single vs many agent relationships? If you’re building this space I’d love to hear from you / try your product .. I’m also curious to hear everyone’s views on the topics above.

Reply with things you want a personal ai assistant to do for you More instinct invites, here t.co/e9BTrmOFNY

So, lots of people have been raving about @noahrshinn's t.co/cFtEr6Qwtt, but when I asked the agent about documentation, it said there wasn't any - but it could answer any questions that I have. Come to find out there is a privacy policy at the root of the Instinct website, but not accessible from the backend where the agent sends you, and where you set up your connections (t.co/mAGSo8ZpEa). It is wild to me that people would connect their work accounts, credit cards, and sensitive login data to a system without a readily accessible privacy policy. It is also fishy to me that the Instinct agent would say there isn't a privacy policy, when there clearly is. It is a cutting-edge AI agent, right? After diving into their privacy policy and terms, I wouldn't give Instinct access to my work inbox under its current policy. Connecting an account can expose conversations with clients, coworkers, and everyone else who trusted you with their information. Disconnecting an integration doesn’t delete what Instinct already copied. Model training is opt-out, with a safety-review exception, and opting out doesn’t undo prior training. Google Workspace API data has explicit protections, but you shouldn’t assume those apply to everything else you connect or upload. And this is the huge one - the terms also let the agent enter binding agreements on your behalf while putting responsibility for unintended actions on you. I work with AI agents. I want them to be useful. But before an agent gets access to client conversations or permission to spend money, I want enforced approval controls, clear retention limits, and deletion I can verify. I wouldn’t connect sensitive accounts until those questions are answered. Read the privacy policy and terms before granting access. And remember, kids. If a product or service is free, you're the product. It is a shame, because the respose times of Instinct are truly amazing, but if you connect your work email, iCloud, and plop your credit card data into a product that has basic HTML as the home page, and a fairly predatory privacy policy - you're a braver person than I.

The killer feature of these new B2C apps (Instinct, Muse, etc.) is to productize browser use in a fast and friendly interface. Most websites, unfortunately, don’t have APIs for agents yet, and their UI is often buggy (e.g airlines, doctor websites) so it’s a delight to delegate that to AI.

instinct > grok bot > muse grok and muse models are not nearly as good as whatever instinct is using. idk what instinct is doing, but it's really incredible.

every tweet about Instinct is some stuff you could've done in codex for months now, what am I missing? t.co/msysZuwqmv

Muse is the best personal agent. And I’ve used them all (OC, Hermes, GrokBot, Instinct). It’s faster, cheaper (free), and easier to use than all the rest. Seems most secure too. Still like having a personal agent on own hardware for certain use-cases and will use each for unique connectors (like GrokBot for X) but Muse will be my main agent at this point. Well done, @alexandr_wang

Muse, Instinct, and your AI agent are about to need an ID. Visa, Mastercard and Alipay are writing the spec. Muse, Grok Bot and Instinct all launched in the last few weeks, and all three will spend your money. Grok Bot, Muse and Instinct all get a single-use card from Stripe for every purchase you approve. Early users describe waking up to things their agent bought overnight. --- This works for now. But not every merchant will work with this. They'll want different wallets. And they'll need to know who the person is behind the agent. They'll need to Know Your Agent. We have some experience in Know Your... [Insert entity here], for example: --> KYC (Know Your Customer): before a bank opens an account, it checks that the person is who they claim to be. --> KYB (Know Your Business): the same check on a company, its directors, and whoever ultimately owns it. --> KYA (Know Your Agent) is the version for software. On Thursday Ant International, Mastercard and Visa said they are building a shared KYA framework. --- Each network had already built its own: - @Visa: Trusted Agent Protocol - @Mastercard: Verifiable Intent - @AlipayPlus: Agentic Mobile Protocol Now they're smooshing those together with a framework that checks: Who operates this? Agents are linked to a validated operator, cardholder, or business. Who certified this? Agents are assessed against security and behavioral requirements. What did it do before? Every agent is scored on identity and transaction signals as it goes, and can lose its certification. --- Baking in ongoing monitoring is SMART. KYC happens once at onboarding (although you should do it "continuously"; many still don't, and how monitoring happens varies wildly). KYA happens on every transaction. Ant's CIO says the signals will include capabilities, behavior and execution performance. That looks like a credit file for agents, but one that updates in real time. A good agent builds a reputation it can carry across networks. A bad one gets locked out of all of them at once. Visa and Mastercard's product is trust between strangers buying and selling. Agents are the new strangers. --- Now the scale. Ant International's Alipay+ connects 150 million merchants to 50+ wallets and 2 billion user accounts. Alipay AI Pay, its AI-native payment product in China, passed 100 million users in February and ran 120 million transactions in a single week over Chinese New Year. The agents launched in San Francisco. The rulebook for trusting them is being written in Singapore, on a platform convened by MAS, with both Western card networks and a wallet network of 2 billion accounts at the same table. In a world that feels more polarized than ever, this is a surprisingly global move from Alipay. This is one of those tiny signals that will be a huge deal in a few years time. Pay attention and you can plan accordingly.

been using instinct for about a week now. it's single handedly the best personal ai assistant i've used. i've used it to book plane tickets, book a dmv test, sent gift cards to family members and much more. the things that stand out are: - communication. perhaps they're not using an off the shelf model? it feels very different than interacting with other models - proactiveness. theres a set of prompts they have the agent run (prepare meeting notes, look for saving money, check-ins). as well as the agent being really good at brining tasks back up. - browser use. i personally haven't been blocked by any site thus far.

I don’t think I’ve ever publicly glazed a model or agent, but Muse agent and Muse Spark 1.3 are phenomenal. They’re now my default personal agent and coding model, I don’t know why I’d use Instinct going forward. Why Muse assistant is goated: 1. Incredibly fast responses The second I click enter I see ward streaming and response, including for complex requests it feels legitimately instant. 2. Top-tier browser use A Concierge an act on those sites faster than I can use it to file an insurance claim and even let me watch and take control of the browser whenever I needed. I have not experienced that with any other agent. 3. Great use of context without being intrusive Just from my conversations, I found that it infer and creates Goals and suggests Ideas for tasks I can use Muse for. Typically this takes me a lot of copy paste, capture and refinement 4. It makes Instagram interesting I barely open Instagram and prefer X, but in the feed provides, it gives me interesting content that I’ll surprised is generated from Instagram posts. I didn’t know there was useful content there. 5. Has all unexpected features, but agent is so useful. I don’t need to use them. Up till now, I have expected features like memory and document artifact management because I’ve learned that agents weren’t useful enough to store information memory like I need them too for the above points even though I can see all my artifacts. I never find the need to open them in Muse 6. Muse Spark 1.3 is the best of Opus and Grok The last few months I found, Opus 5 to be to be very intelligent, but too long-winded, while Grok 4.6 was direct and effective, but not very insightful. Spark 1.3 is is insightful, intelligent, and direct in coding and non-coding use cases. No more choosing between depth and speed 7. Price. Despite being uber powerful Muse Spark is cheaper than any comparative model on Open Router ($1.5/M vs $5/M input for Opus) So naturally , I promoted it to my iOS dock Only call out so far: on image gen I find Gemini/nano banana matches my intent better, and is easier to prompt. (To be clear this is my opinion as an AI nerd and builder, not as a Meta employee)

Instinct is cool...BUT.... it uses shared iMessage lines, which means many users are talking to their agents on the same number. With @inkbox_ai I gave my Instinct a dedicated phone number that only belongs to it! Now it can call and text anyone!

The fastest way to cement yourself as an absolute weirdo in Europe is by using Instinct. Instinct reached out to my BnB to try to find a cab service in a tiny Belgian town. Cue confusion by the sweet woman at the desk. She then reached out to my French friend who booked the reservation, saying that someone by the name of Instinct is claiming to be my AI assistant and making asks of her. This then led to my French friend explaining that I was American (implied: absolute weirdo). He then did the job that Instinct was supposed to do, and also apologized. She remained confused. Net: befuddled woman, annoyed friend, and happy me who had a cab booked.

Okay this is cool. Unprompted - Instinct found a $245 settlement credit card I never activated. It reached out to me! I didn’t tell it to monitor my inbox to find money.

the more i use muse, the more i realize that it's the single best in-house product meta has brought to market since facebook. showing the progress realtime with browsers is so key. instinct has some limitation with it since it doesn't have its own native app.

the biggest issue with trying to use all these personal assistants is having to reimport all of my context every time some of it can be made easier with shared storage (e.g. my file system), some tools have migrate functionality (e.g. codex import), but it's still a large amount of work i've tried/am trying codex, claude code, claude cowork, grok bot, instinct, muse....i ultimately want one assistant to rule them all

i asked instinct to move my united flight for me it rebooked me on a nonstop & even got the change fee waived the last time i called united i heard the same eight bars of hold music for hours

Instinct killing 2FA is the “move fast, break things” version of AI startups. For AI agents there’s a trade-off between UX and cybersecurity. 2FA gets in the way of the assistant to be effective. Instinct is willing to take the security risk of eliminating 2FA to get there (if users get hacked they can take a reputation hit). Taking this risk might actually turn up to be a massive moat - Meta (and their assistant Muse) or other corporations can’t make the same trade-off and take anywhere close these levels of risk. Choosing UX over security this aggressively is something only a startup can do.

Ok, @Muse is amazing! I've been putting off getting it set up, but now that I have I'm blown away by how well it works. It's just all the attention to the finer details, so good! The only thing I'm missing from Instinct is it's proactiveness. Try Muse out if you haven't yet, you won't regret it!

Paul Klein (@pk_iv), Founder & CEO of @browserbase, reveals the top AI agent use cases: “The most surprising deployment of agents that we've seen so far is often agents that are working on legacy workflows.” He also referenced Instinct, one of the most talked-about products building agents that use a browser: “We've seen Instinct do a lot of amazing stuff, really pushing forward what's possible.” The use cases are increasingly everyday tasks: › Navigating legacy websites & workflows › Monitoring flights for changes › Checking schedules & updates › Moving PDFs between websites › Uploading expenses › Filling out online forms › Making restaurant reservations › Handling repetitive browser tasks › Doing work you’d normally delegate to someone on your team “We look at these screens every day to monitor things. I think AI's gonna take that over.”

Went to LA last month and forgot my laptop at home. Forced me to run my whole business from my phone 2 biggest limitations: 1. Video calls with clients are kind of jank from your phone. More professional to take them from your laptop webcam 2. Wispr flow just more annoying to use on your phone Besides that, wasn’t too difficult. And now with instinct can sign documents and have it draft longer emails which makes phone only even easier

Frontier AI won't slow down. Nobody slows down unilaterally… and coordinating is illegal without an antitrust waiver. However, let's say it did. What are the implications? *1) Does it reduce demand for AI?* In the short to medium term: no. Most companies are still wrestling to figure out how to implement it, with a great ROI. The frontier has run much faster than the use cases so far. You can solve extremely complex math problems with the best models - yet we are only just starting to have useful AI assistants. Beyond consumer search, coding and perhaps law, penetration is still low. Models are good enough to process insurance claims, customer service calls, and dozens of other use cases. These do not need a better model. However, they need cheaper tokens and someone to build the workflow. If anything a pause is great for the app layer. There would be less talk of "the next model eats my startup" - and the labs would need to start competing on price and distribution instead of capability. And of course, truly useful consumer AI will be a massive source of demand too (muse, instinct, town etc…) *2) Do we need any less compute?* I spoke to a team trying to secure 1k GB200s on Friday… well even with tens of millions in hand, it's not easy. Demand vastly outpaces supply already. Pacing, in Dario's own words, does not mean stopping training. It means slowing capability gains. The compute keeps running. But it will move: more inference, more reinforcement learning, more world models and robotics. Great occasion for adjacent labs to catch-up / overtake too. And given 1), inference demand compounds regardless. So no. *3) Will open source overtake the frontier?* Well, that is the most interesting one. If open source truly innovates (instead of distilling and copying), then it should. But open source today mostly means Chinese labs. And every big open leap has landed months after the equivalent frontier leap. Never before. DeepSeek has real innovation on the training side (MLA, FP8). But a post-training gap closing so fast is hard to explain without distillation, and Dario clearly agrees (and spoke about this many times): his essay pairs pacing with a crackdown on distillation and weight theft. So: open weights become the floor in terms of speed. The frontier can only pace as much as China lets it: the stakes are too high to fall behind. *Conclusion: if it did, I would remain bullish apps and inference, but turn more bearish lab pricing power. Which is probably why it won't happen.*

Apple or Google is going to buy instinct if they keep/get consumer mindshare imo But outside X most people I know don’t know what Instinct is t.co/9lzmwYiumH

Can someone invite me to instinct pls

Muse shows how much better a full new agent app experience is vs bolting on the iMessage. Incumbents still have distribution advantages as well. I’m sure Instinct knew this but chose low friction over better experience and it will hurt them in the long run.

Instinct is fucking unreal man and such a relatively simple tool. Even if models are "pacing" themselves, I strongly believe that 99% of participants are entirely unaware how good AI has gotten.

I’ve been using Instinct, Grok Bot and Muse and comparing output against similar inputs. Here are my thoughts:

Apparently I have 20 Instinct AI invites to give away that expire in 48 hours.

Anyone got invite for instinct? I swear its for educational purpose only

instinct invite incase anyone wants t.co/LILUvjjgiZ

muse, instinct, poke, etc. are converging in terms of capabilities i think we will see "personality hire" type of decision-making as capabilities inevitably fully converge in other words, we'll go from "tool" to "teammate" very soon

Someone just told me *Instinct* recommended they should attend Dirty Jobs as a top physical AI event for 2026 as their reason for applying Agentic recommendations are the (very near) future. Ask your Instinct if Dirty Jobs is a top physical AI event you should attend and LMK what it says?

Ok I’m down to try instinct. Who has an invite?

here are 15 random-est things i've asked instinct to do and it outshined every time 1. got me an API key, live-tested it, sent integration notes 2. registered me for 11 founder events + polished my luma profile 3. sorted my US SIM plan the minute i landed 4. dogfooded vaaya's signup, sent a good/bad/ugly audit 5. mined my beta ledger and slightly corrected the traction numbers i'd been quoting 6. filled out some key accelerator application forms 7. logged in on vaaya with instinct email id and started churning out legendary videos 8. checked me in the second the window opened, free aisle seats on all 4 legs to SF 9. got traction numbers across socials and who interacted with our posts and how to engage them further on DMs 10. looked up my two dream hires, wrote the angle of approach 11. found 5 influencers for our offshoot, with the play order and the simple pitch 12. found demo storefronts hiding in my catalog, traced to one operator 13. verified a vendor's rollover terms before i spent $49/mo 14. reads my inbox, tells me what's real vs bank spam 15. settled the star wars watch order. (ep IV obviously) on a message-only internet pack on flight and what else is worth watching having a chief of staff that never sleeps hits different. can someone share the muse invite been wanting to test it for a while

This week my AI chief of staff (his name is Otto) sent me a prep five minutes before every external call - who's in the room, why I'm there, the angle, and the last thing on the thread. It flagged a bill the minute it landed and told me which kid needs what for school this week. And this morning it traded founder referrals with another investor's agent while I drank my coffee. The last part is new. Agents can now work with other agents, each one staying inside its person's permissions. Mine already coordinates with my wife's. All I did was say yes. I have 15 Instinct invites, good for the next 48 hours. Send me the best thing an agent has done for you - or the one thing you wish yours would do - and I'll pick my favorites. P.S. My agent also wrote this post.

During my 5 hour road trip today, Instinct AI: Found two investment ideas that fit my parameters and look primed to rip. Cleaned up and labeled emails in my personal inbox. Negotiated a lower TV bill through live chat with my provider. I drove. It paid for some of the trip. @noahrshinn

everyone: using instinct to get day of resys to the best resturants my girlfriend: refreshed the amc website 10x 6 weeks ago to get us 10pm Sunday tickets to 70mm odyssey

OpenClaw is for the prosumers. Instinct is for the early (connected) adopters. Muse is for the masses. (Now, how do we spin-out Muse from Meta? Asking for a friend ;-))

Anyone have an Instinct invite link they can share with me? Many thanks!

All of my thoughts on Instinct. Cool agent platform, with a terrible privacy start. t.co/dD62Vsbm7R

“…woman at the desk. She then reached out to my French friend who booked the reservation, saying that someone by the name of Instinct is claiming to be my AI assistant...” Product failure tbh No one would be weirded out if it just called itself “Bob” or “Alice” t.co/S7Lftmw1qE

apparently @noahrshinn's instinct uses kimi k2.6. it resisted pretty hard when i tried to get it to reveal its architecture, but here are a few things i found: - uses ephemeral sandboxes rather than persistent ones, unlike grok bot and muse - relies heavily on residential proxies - it can dynamically create secret vault entries for you to populate through the web app. they claim ai doesn't see secret but it technically can. - token limits are daily and suspiciously high right now, at least 100m+ - their rl-tuned model seems heavily optimized for browser use - so far, it’s the only multiplayer ai i’ve seen that lets your bot talk directly to your friends’ bots through a trusted-person network it's a good product but is it 10x better than @bot or @Muse to beat their distribution? i'm afraid not.

What I actually did with Instinct this weekend: - Enriched 90 event registrations with company, size and funding data, flagged who's a real prospect vs someone trying to sell me something - Sent 83 personal confirmation emails for our Boston happy hour. Individually, not a mail merge - Built an entire NYC campaign as a dry run: venue shortlist with walking distances, 34-person invite list with sources, email copy, sequence loaded as a draft. Nothing sent, everything ready - Deleted 9 dead email senders and caught 4 more that broke overnight - Chased down a refund I'd forgotten about - Pulled my messy Notion idea list into one clean doc All of this over text, mostly after 10pm.