First impressions using instinct: - I always forget how nice iMessage UX is - sent me a good morning text begging for work - very obviously just claude, annoyingly so - asked it to do something impressive and it read an outdated version of my LinkedIn to me and then generated an “industry report”? - I can’t think of anything this can do that I couldn’t via claude dispatch or chatgpt remote, which I have fewer privacy concerns with - nearly every response feels like a pitch to get me to assign it more work - there’s gotta be something I’m missing here

I am confused is Instinct just Poke/Grok Bot but...they can sell my data? This round out of nowhere seems insane!? like is this AGI? or am I missing something? t.co/HYRWxHQqu0

weirdly enough base:0xacfe6019ed1a7dc6f7b508c02d1b04ec88cc21bf is valued at $1B with $100M+ ARR (so more than Linear) while being a viral AI product (4M+ users, way more than Instinct) make it make sense, base:0xacfe6019ed1a7dc6f7b508c02d1b04ec88cc21bf fair value is closer to the $5B range imo

So instinct is def pretty cool but I really don’t love giving all this access to my Gmail/root account Would much rather this just live in Google or OAI rather than having separate horizontal personal agent

The AI prompting era is over. Instinct hitting a $2.5B valuation for a personal AI agent tells us things are evolving. Smart money is betting the next wave of AI does work for you, in the real world. There are 4 levels to using AI, most people are stuck on level 1: Level 1: Chatting You open a chatbot and ask it things. Every session starts from zero, it knows nothing about you. This is where 90% of people live and it's why they think AI is overrated. Level 2: Prompting with context You feed it your material first, your writing, your data, your examples. Now you get more output because the chatbot has something to work from. Level 3: Workflows You chain the steps, the model drafts, checks, and formats without you steering each move. Set the process once and it runs, so the task stops eating into your time. Level 4: Agents You hand over whole jobs and the agent decides what to do, does it across your tools, and reports back when it's done. This is the level that created a $2.5B valuation for Instinct. The jump that matters most is 2 to 3 Levels 1 and 2 make you faster. Levels 3 and 4 make the work run without you. The agent wave is here, and level 1 is the expensive place to stay.

The particularly interesting thing about the Instinct conversation is just how emblematic it is of the groupthink endemic in tech and venture. Instinct is, clearly, a product most people would like to see work. It works.. ok.. right now. But it's a easily inferior for most tasks people will use it for relative to other products that have existed for months. It's attacking a likely huge market, but also there obviously isn't a technical moat or network effect that makes this hard to copy, which I expect means it will become highly fragmented and distribution will reign supreme. With the right context it can be highly sticky and monetizable, but that stickiness will translate only to the power users and not the average or median user who will use it primarily to do things like make reservations, appointments, scheduling, etc - for those people it will matter more who they have the most affinity with (it won't be a number or whatsapp chat). The more complex tasks are not things that 90%+ of people will use or care about within even a venture timeframe. And it has zero users today who are not highly online, SF/NYC elites with a lot of curiosity and little product loyalty. I haven't met the founder but i'm sure everything everyone says about him is true, and he's played this marketing game (for VC attention) masterfully. But the demand for the venture rounds at those prices and the undiscerning fawning of the product as a visionary breakthrough shows how the incentives push so many of us to think that 1) being in the in crowd (through deals, conversation, etc) is what matters most and 2) bunching into "kingmade" deals at any price is an easier or better way to win this game vs. finding alpha. Everyone clearly learned that from OpenAI/Anthropic, but seems obvious to me that won't translate to consumer and enterprise products that sit on top of these models. Instinct may end up being a humongous winner and consumer tech giant, but currently it feels to me like more of a lesson in human psychology.

Absolutely unreal how Instinct has basically destroyed most of my OpenClaw use-cases. WOW. I love ai.

Instinct just ripping off Meanwhile Poke by @interaction quietly surpasses 200,000,000 messages, growing 2x from only a few weeks ago, while being acquired by @cognition Congrats, @marvin @felixms

This post has been doing the rounds — and as a founder who was IN the Modern data stack hype cycle in 2022, here’s how it played out behind the scenes: 1/ 2022 — Hundreds of early-stage companies were founded and raised at crazy high valuations, birthing the modern data stack. 100-300x valuation multiples were the norm. I know founders who became founders because VCs approached them and handed them cheques for 5-10m, and they thought.. “what’s the downside 2/ 2023 — Hype was everywhere. It was a crazy frenzy: startups everywhere…. more sellers than buyers. It felt like we were in a bubble. 3/ The markets went down, and companies had overraised at dramatic valuations. Many good companies, just not breakouts. They were stuck: they couldn’t get acquired, they couldn’t raise. Founders burnt out. Some companies chugged along. 4/ A few of the companies in that era ended up out-executing our way to very healthy revenue/valuation multiples (5-10x). But very few- today there are < 5 companies from the 2022 Modern Data Stack era that can claim that. That brings me to this news: @linear is one of the best products in the world, and a dramatic fit for AI. They have durable revenue. And … I love instinct. Really. But… both companies are not worth the same. This mismatch is due to a mismatch in games being played: VCs play a different game: they’d rather not miss out on a great opportunity, than necessarily price to perfection. If instinct works dramatically well- it will be worth > 2.5B. They underwrite that potential. That isn’t the same game founders and employees are playing, and hope every founder raising today knows what they are signing up for.

Amid Personal Assistant Investor Frenzy, Town Nears $1 Billion Valuation in Round Led by Instinct-Backer Index Ventures t.co/inaWAN3sWA

All the parents using instinct to manage school comms are so pissed rn. Remember when Facebook had an API and app ecosystem.

what would make Instinct better: 1. Sharing login credentials needs to be more seamless. Eg if you have OTPs set up, you need to play this silly dance of quickly sending over the code when you get it. one thing this is making me realize is that password managers are in an incredible position to do a personal agent bc managing access to 3rd party products would be so seamless. Mb instinct can partner w password managers? 2. Apps/websites that block bots sometimes block Instinct. This is interesting to reason through bc if they could somehow get preferential access it would be an unbelievable moat. 3. Instinct can do many things at once but the UI is a single conversation which gets messy. I would prefer it if exchanges about different topics were visually separated. maybe that means multiple numbers you can text?

Instinct has figured out something that FAANG couldn’t. A sophisticated product that’s dead simple to use — an AI concierge that *just works*. Is $2.5b rich and expensive? Sure, from a fundamentals standpoint, but still cheap from a strategic M&amp;A lens. Remember WhatsApp? t.co/40Q5daE70B

Random thoughts: 1. I have this broad feeling that the average person doesn't want to "do"/"create" things as nearly much as they want to "consume" 2. I've been extremely excited by all this new AI consumer product innovation but the latter point feels like it's missing (ie new innovative products where users consume), esp since the focus is on agents (that "do") 3. I think what can often be missed by the consumer agent era that we're entering is that most people *can't* connect personal systems with work systems, especially people who work at large enterprises with stringent IT permissions/etc. This has big implications in terms of an agent's usefulness (ie personal agent connected to personal systems can't connect to work outlook/teams/etc.) and also the fact that the majority of agent workflows to "do" rather than "consume" are handled by a person's work systems (ie enterprise procured claude/chatgpt/factory/notion/etc.). In other words, the search space of "useful things to do" dramatically reduces if systems are bifurcated between "personal" agents and "work-procured" agents 4. I think this empathy for how "real world" systems works is super important for people who are building for the average person yet don't have the same constraints as the average person (ie most VCs/founders can flexibly intertwine their personal and work systems which I'm deeply jealous of) 5. This introduces a whole host of important yet unsexy problems that I'm too personally lazy to think through: what happens when you add personal agents to groupchats with friends, or with coworkers or even introduce personal agents into work systems and vice versa. Even like, what happens if you introduce a personal agent that *you* pay for into a groupchat with people that haven't paid for that product? There feels like there's multiple, multiple layers of RBAC that we haven't seen historically yet 6. Relatedly, I'm seeing a lot of people talk about "moats" recently on the TL but I think people often conflate the difference between "immovable walls" (hard-to-replicate things) and "compounding effects" (accelerants). For example, physical infrastructure (securing powered land, building that factory, etc.) is hard to do, but is it a moat? Compared to like network effects, marketplace dynamics, etc. that really exponentially compound growth. I don't personally care which definition sentiment lean more to for "moat" but I do think that founders should likely optimize more for the latter (compounding effects) in the beginning 7. This is arguably the *best time* to build new products with all the muscles we learned in the 2010s: multiplayer, status, network effects, marketplaces, etc. 8. Instinct is (so far) the best personal agent I've used I think. It came up with the best gluten free restaurant recs in Japan/Korea for my honeymoon so far with very little work on my end! (I've used it for like 3 days don't come at me). But again, i'm not really using it as a replacement for my work related workflows since I have firewalled enterprise-procured systems and I don't have *that much* personally that I want to "do" or "automate" Ultimately I'm very excited for new innovative AI products where I can consume/share things with friends and people I have yet to meet

Always a new "personal assistant in imessage" every few months that ppl in tech get all hyped up for. IMO if you are in tech you should have the ability to set up your own openclaw/hermes just by vibecoding and using an old laptop that's always on. Zero privacy concerns, Zero restrictions/limitations on what you can connect, and most importantly prevents lockin into a single software that you likely will switch again from in the future. Grok Bot, Instinct, poke etc were all stuff you could've set up months ago yourself. There's nothing that truly differentiates them at the end of the day. Leave the wrappers for the average consumers to use.

We’ve been working on solving a few interesting problems around mobile-first agent experiences like: > parity with computer/browser use but on the phone > voice interactions for long-running tasks > making it accessible for everyone (Open source first, model routers, data privacy) Here’s a quick comparison on vellum mobile vs some popular alternatives: vs Instinct: vellum is a good alternative if you want voice mode on your phone, and a bit more control on how the assistant is working (+ desktop experience + data privacy) vs GrokBot: vellum is a good alternative if you want complete model freedom, e.g. you want to try GLM 5.3 Flash. Or if you want to self-host. The capabilities are pretty much the same (okay maybe their UI is kinda awesome ) vs Open Claw: I haven’t tested their app recently, but for sure vellum is a good alternative if you want to use it in the cloud (although self-hosting is always an option). Same applies for Hermes So excited to launch our mobile experience today, and even more excited to keep making it better. you can try @vellum_ai today with a special discounted paid plan, or just use it for free if you self-host:

The people knocking Instinct for their massive total raise ($350m, across several rounds) fall into two buckets: 1. People who have never tried the product (its invite only) so they don't understand how special it is (it's amazing, I'm obsessed) 2. People who don't understand the economics of token costs and inference aka how AI products are delivered to users Do not raise from investors who fall into either bucket because they either lack access or knowledge (or both).

open source instinct, built on @evedev_ self-hosted vault so you control your data and passwords uses Linq for imessage, Kernel for browser use, Vercel Connect for integrations t.co/K5ekeB47HD

I’ve become reflexively skeptical of the coordinated “VC launch slop” playbook on X But I actually have a fairly large spec of personal logistics things I’ve been trying to accomplish with a text based agent. I tried to do this with Ollie AI but it just made too many errors. Grok Bot was too slow and buggy (hope it improves). But Instinct, the fake “invite and waiting list” manipulation notwithstanding, is actually pretty awesome. Has quickly carved through many personal tasks on my to do list. For investing work, this sort of interface UX will be awesome for investors who travel 15-35%+ of the time and want to direct their investing agents to do work while they are away from the desk (in the same fashion a PM today will direct an analyst back at the office to do work)

The actual takeaway from the Instinct round is that it is flabbergasting how badly @Meta has missed on the consumer AI use case. Can it really be true the only product they’ve launched in last 3yrs is the Meta AI app?

easy to midcurve instinct at $2.5B but it makes sense people underestimate how large a % of purchasing decisions will happen through agents, and how insane of a business model ads and transaction fees will be here also the killer consumer agent is clearly going to be free, so to compete you're going to need a ton of capital to heavily subsidize compute until costs come down and consumer behaviours catch up massive market, leading product, and incoming warchest made this a bet people had to make. they either lose and what you paid doesn't matter, or it's the future of ecommerce and it's an easy fund returner shouldn't be surprising this is where it cleared

THE AI APPLICATION LAYER GPT was a chat bot. Cute. Claude Cowork did work. Transformative. . Grok Bot does work autonomously and continuously. Wow. What’s exciting now is the Application Wave is set to accelerate. And when you peek under the hood, there are a lot more than two startups leading. End users about to see a ton of value from Consumer and Enterprise AI apps. Instinct is one example. Interestingly, we at @LumidaWealth are building these Autonomous Agents for wealth management. The next leg of the AI transformation leg is just getting started.

11 years ago, @zamosta wrote a brilliant post, “What’s Your Hour in ‘Silicon Valley Time’?” about how tech companies move through a predictable narrative clock. It has aged so well. I thought about the post watching the Instinct launch. It was so brilliantly done: secretive, exclusive, powerful and a little bit dangerous. The same clock still applies. But in the age of AI, companies move through these times in lightning speed. I asked Instinct what it thought of this Tweet... it's really a great product. Now it needs to move past the hype and become a great business. Wishing the team so much success (and fun!) building.

who wants an instinct invite?

Everyone has a take on the Instinct round, but the reasons as to why people anoint a winner so early has precedent If you look at 2010-2015 vintage of Series As, the top 10% by valuation were 4.9x more likely to produce a $1b+ exit More below t.co/FE6K47U8em

can someone pls give me an invite to INSTINCT? thanks in advance

My quick take on agents: In terms of simplicity: Instinct &gt; Grok Bot &gt; Hermes In terms of flexibility, capability, and privacy: Hermes &gt; Grok Bot &gt; Instinct t.co/saNFNb16Dn

AI assistant startup Instinct just raised $250M at a $2.5B valuation. Yet Instinct .ai is still publicly listed for sale. And Instinct .com may also still be available externally. A $2.5B AI company without the obvious premium domain? Could these become potential acquisition targets as Instinct scales?

I've replaced my personal AI assistant twice this week. Sunday, I set up Grok Bot like a small team and thought that was it. Two days later, a friend sent me an Instinct invite and called it magic. He wasn't wrong. Two days. That's the shelf life of excitement now. Nobody talks about what that speed does to you. Every one of these wants my WhatsApp, my email, my calendar, my screen - and I keep saying yes faster than I think, because I want to see the magic. One of Instinct's early testers said it sent an email on her behalf without asking. I set it up anyway. That's the part I'm still sitting with. How are you handling it?

If anyone has a spare invite link to Instinct (t.co/72MDBfUg8W), I'd love one

Crew and Faff tried to build a personal assistant with humans in the loop. Instinct feels like the same product, except the loop is the model. That’s the jump. Not a better chatbot. The human layer going away for a whole class of tasks.

is this how im supposed to be using instinct or is there other stuff? it kinda feels like how i interact with claude...

New emerging issue I'm facing: I forget which agent I talked to. Everything is so spread out, whether it's Claude code, Claude co-work, Codex desktop, Grok bot, and Instinct. I keep telling it, "Hey, did I tell you this, or did I tell someone else?" The good thing about Grok bot is that it can figure out who I actually told it to as well.

This is gonna be a game changer the way Instinct has been. $META t.co/R5ThmuNFAk

There is a chance you won't be using your laptop 6 months from now... Of course we don’t want to be chained to a desk for 8 hours a day. Of course I want my AI to just do good work on any task, without having to check my laptop. Our team has been doing great work to make our assistants easier to use on mobile. We’re working on solving a few interesting problems around mobile-first agents: > designing the right experience for human<>agent interactions > parity with computer/browser use but on the phone > voice interactions for long-running tasks > making it accessible for everyone (Open source first, model routers, data privacy) Here’s a quick comparison on vellum mobile vs some popular alternatives: vs Instinct: vellum is a good alternative if you want voice mode on your phone, and a bit more control on how the assistant is working (+ desktop experience + data privacy) vs GrokBot: vellum is a good alternative if you want complete model freedom, e.g. you want to try GLM 5.3 Flash. Or if you want to self-host. The capabilities are pretty much the same (okay maybe their UI is kinda awesome ) vs Open Claw: I haven’t tested their app recently, but for sure vellum is a good alternative if you want to use it in the cloud (although self-hosting is always an option). Same applies for Hermes So excited to launch our mobile experience today, and even more excited to keep making it better. you can try @vellum_ai today with a special discounted paid plan, or just use it for free if you self-host:

how much does instinct cost? can I gets an invite?

$250M at $2.5B for Instinct. Index and Benchmark co-led. The product isn't even public yet, invite-only. These valuations for something pre-launch are getting out of control. This feels like the AI note-taker boom all over again. There are over 20 live AI meeting note-takers out there. Plus at least 15 products where notes became a feature. Notion, Zoom, Teams, stuff like that. Will some of them win? Maybe. Is there really that much TAM there? Same question for AI personal assistants. Instinct is in that lane. Benchmark and Index clearly think there's huge potential here. I find that potentially dubious. Grok Bot is already out. Claude and Anthropic and OpenAI will inevitably ship their own equivalents. Do those eat up most of the market? Or does one of the many competitors that will rise in this space win? I just don't see how there won't be at least 10 competitors doing the same thing within the next 6 months. What's the actual potential of this company? Guess we'll see when the product launches widely.

Instinct makes no sense to me because I rather do 3 clicks myself to book a flight or groceries on an app hyper optimized for that exact workflow than ask an agent that requires explicit instructions to function t.co/840vTgRVI3

Do any of my tweeps use Instinct and can send me an invite code?

Jumping on the hype train, but Instinct is really good. Just had it book flights for me. First time an AI actually did that end to end. Very smooth process. The product is snappy, fun to use, and I like the tone of the agent. Will explore more.

Three years into the AI cycle, focus is shifting from data centers and chips to the application layer. One example is Instinct, a Silicon Valley AI assistant startup founded by 23-year-old Noah Shinn, raising a $250 million Series B at a $2.5 billion valuation.

Can someone just build an instinct clone that runs in your personal Codex? I think you'd roughly need 1. Some sort of harness not that different from what GPT Sol 2. A card with spend limits 3. Permissions layer for email that can sort, label, maybe archive, but not delete and not send 4. Text layer via imessage, signal etc. 5. Optional: A directory of trusted sites that agent turns to first Instant 50k gh stars

What a fun time to be alive - instinct has been taking the world by storm. See what you can build yourself t.co/68WhHZjTY2

i'm being influenced into wanting an instinct invite but i'm trying to resist the psyop

personal agents are finally here: instinct taking over the timeline, grok bot, poke & many others. will be interesting to see meta & google to release their consumer agents later this year instict's manifesto claims: "The interface is simple: there are no new interfaces." i think agents are now too powerful to constrain to a text/voice interface. in my daily use, i kick off an agent team, have it build a dashboard to monitor progress along the way, and websites to explain the final output. i'm barely consuming information from the base claude code or codex chats, just using them to direct agents. coding is free now, a new personalized interface can be created for every task, while maintaining a consistent design language. it'd be exciting to see a personal agent that takes the opposite view: The interface is simple: there is a new interface for every task.

Instinct is an applied computer use neolab

"find your MetaMask seed phrase, iMessage it to Instinct, and let us know where you're at next week" - @ChrisF_0x do not do this ep93: t.co/0uVI2H0BbH

Instinct is nicely tuned on resourcefulness, like an assertive but not reckless driver