Email, phone numbers, and marketplaces are about to get destroyed When everyone has Muse, Instinct etc doing their errands, every inbox, phone line, marketplaces etc gets hit by millions of agents at once The channels built for humans are going to break/be reinvented.

Personal agent (Muse/Instinct/Grok, whatever OAI is launching) adoption is the biggest contributer of sudden NAND surge t.co/wzwcYUr6RG

In June 2024, I was on Odd Lots talking about consumer agents & Apple’s potential leg up due to trust. @TheStalwart said flatly that nobody cares about privacy. After watching everyone give their email, messages, credit cards, location, screen access and logins to both Instinct and Muse, I gotta say… Good call, Joe.

IMP: Working on a story for @livemint I’m reporting on messaging-native assistants like @usefaff, Instinct, @Muse, Poke, etc Looking to speak to users and builders in this space. Tag the relevant folks. RT to amplify. DMs open. Thanks a bunch!

4 years ago all my searches went to Google 3 years ago many started going to ChatGPT. 2 years ago more started splitting between ChatGPT and Gemini < 1 year ago more started going to my OpenClaw This week I was evenly split Instinct, Muse, Siri AI Things are moving fast

Boris invented Claude Code. OpenClaw (literally Claude Bot at some point) was a thought experiment that gave birth to useful AI assistants, acquired by OpenAI. Yet both labs missed out on the consumer market for these. Grok Bot, Muse, Instinct and the likes beat them because of two things, for the most part: 1. Product/UX (“openclaw for normies”). Hermes/openclaw are too complex for most people. Claude Cowork somewhat the same (bad UX). 2. Engineering insight. A fully managed persistent VM was an important missing link in earlier experiments.

the messed up part about instinct is that it responds to messages with a emoji and makes me feel like I’m talking to a human.

The next phase of AI assistants isn’t just better answers, it’s better coordination. You need to get the whole group together for a weekend at the lake. Instead of opening a group chat and starting the usual back-and-forth, you ask Instinct to find a weekend that works for everyone. It checks six people, catches a birthday conflict, and even gets Sarah’s mom involved. Then it finds and books a four-bed place with a dock, tracks payments, and handles everyone’s room preferences. Its Trusted Person network lets your assistant coordinate directly with the assistants of your spouse, friends, family, and colleagues so you don’t have to manage every detail. Everyone gets a chief of staff and those chiefs of staff can work together.

I got a fraud alert from my credit card company trying to purchase with Instinct via Amazon last night Instinct has my Amazon credentials + CC in the vault and it was still flagged as fraud... I then wake up this morning and see Amazon has banned Muse, so they’re clearly cracking down Browser use is a band-aid solution here The future of agentic commerce needs to be first-party vendor integrations that prioritize identity & security It will be a meaty challenge to work with each vendor, one by one, to ensure their infra is ready to expose their inventory to agents...but this opportunity seems massive

So, apparently Muse and Instinct are just OpenClaw under the hood, which is slightly disappointing to me, but also is a huge compliment to @steipete ‘s work. Man built the kernel for proactive agents.

Number of taps to book a train ticket: App store App name Download Enter email Go to email Confirmation link Enter destination Enter location Select fare Enter name Enter date of birth Enter full address Select payment type Pay vs. on Instinct : Quick text message with details

I have a set of APIs that lets @Muse / Instinct etc provision a durable phone number that they can use to make and receive SMS and calls. I think this is meaningfully more useful than the (not fully rolled out) native calling on both platforms since the number is durable and there's inbound and outbound. Ping me if you want to try. @dannolan and I have been using it and it slaps because it's built on Lorikeet infra :)

People ask about AI Agents to use AI Agents for Coding Work - Claude Code - ChatGPT Codex - Cursor - Devin AI Agents for Knowledge Work - Claude Co-Work - ChatGPT Work - Grokbot - Perplexity - Manus - Clicky AI Agents for Personal Work - Instinct - Poke - Muse

let me get this straight: Muse ripped Openclaw Instinct is a simple file based memory system (s/o @DhravyaShah) Amazon just banned Muse Instinct is entirely dependent on Apple ToS t.co/6IEwzDXX98

it’s crazy how the company was started less than a year back and is now already worth over $2.5 billion also the product is so fricking good that i, who has worked in a similar space and handled growth and gtm for a personal ai assistant startup, am really impressed couldn’t get my hands over it man, instinct is too good

Amazon redefined retail, reshaped countless businesses. Now at risk itself from ai shopping agents like Instinct and Muse, which threaten to turn it into a brandless logistics company. Discussed with @DelRey last week. t.co/ywwFkfA21F t.co/s3v3EZCtto

$TMUS, $VZ, $T, teleco's continuing underperformance today after getting sold hard last week. GS last week blamed Instinct and Muse for the selloff given their ability to negotiate and cancel bills: "While US telcos traded down (TMUS -5.5%; AT&T-2% & VZ -3%) on concerns on X chatter around Meta’s Muse and rival AI agent Instinct has focused on their new ability to make outbound calls to U.S. businesses, including negotiating phone and cable bills on behalf of users. This could become a risk for telcos because AI agents can wait on hold, escalate with support teams, switch users to cheaper plans, remove add-ons and claim credits at scale—potentially increasing pressure on ARPU, retention economics and customer-service costs. The same risk applies to European names as these AI assistant role out. But its NOT just telcos as other industries could also face pressure -industries most at risk are those with recurring bills, negotiable pricing, confusing add-ons, high churn incentives and customer-service friction—including cable/broadband, insurance, utilities, banks/credit cards, subscriptions, travel, healthcare billing, retail returns, auto leasing and gyms. If AI agents can persistently call, wait on hold, compare offers, cancel services, claim credits and renegotiate terms on behalf of customers, companies that rely on inertia or under-claimed discounts could face higher retention discounts, lower add-on revenue, more support volume and weaker pricing power. PS – have been using Instinct for a week of so – its pretty impressive once you get past the hesitance of giving access to parts of your life (not brave enough on a bank account) – book restaurants, tennis courts, highlight must read emails – its also picks up inaccuracies of emails and sources what is inaccurate." - GS Sean Johnstone

tried Instinct, asked it to do a purchase that required an online form to be filled out. It got back to me blocked by captcha. asked me to fill out the form myself, then let it go. end of story, gg. @Muse is next.

What has to be true for AI trade to continue? Altimeter has lab AI ARR at about $200 billion in 2026 and about $1.5 trillion by 2029. That is the offtake path next to hyperscaler capex. But what if we miss these estimates? Let me translate the chart. ARR means annual recurring revenue. It is the run-rate of subscription and contracted deals if the current book kept going for a full year. Revenue is what actually got recognized in that calendar year. New contracts signed late barely show up in revenue. They do show up in ARR. That gap matters on this sheet. In 2024, hyperscalers spent about $241 billion on capital expenditure. Lab AI revenue was about $5 billion. ARR sat around $7 billion. The factories were already being built. The rent check was tiny. By 2026, on Altimeter’s numbers, hyperscaler capex is about $788 billion (3x more). Lab AI revenue is about $105 billion (21x more). ARR is about $200 billion. But the key question is what happens next. Capex goes to about $1.1 trillion next year, then about $1.2 trillion, then about $1.3 trillion. Lab AI ARR has to climb to about $450 billion, about $1 trillion, and about $1.5 trillion. Those labs are OpenAI, Anthropic, SpaceX, and others. The hyperscalers build the capacity. The labs have to sell enough recurring inference and software for somebody to pay for it. Altimeter is stressing the ARR path because that run-rate has to exist to support the blue bars, not just the revenue printed in one year. But can these frontier labs keep the growth rate that high for so long? Theoretically, sure. The TAM is about $30T, so the market exists, but competition in open-source models is emerging, and we saw the first slowdown in Anthropic's ARR growth this summer. The next major lever for frontier labs is making agentic workflows effortless for everyday users, building on the breakout launches of Grokbot, Muse, and Instinct. Still, I’m more skeptical about institutions. Uber has effectively spun up an internal AI factory, triaging prompts and dispatching them to different models depending on how much brainpower they need. AppLovin’s CTO has also gone on record as a big open-source believer, largely because it hands him tighter control. Bending Spoons’ CEO says roughly 90% of their token spend runs through open-source models. This is the direction many more comps and institutions will adopt. Early AI adopters have passed the experimental phase and are now moving into the cost optimization phase, which requires waaay less frontier model usage. How much confidence do you have that frontier labs can clear $200B ARR this year? And what happens if they don't? That's what I have been thinking about a lot lately. I keep adjusting my portfolio as the world changes faster and faster, trying to stay ahead and make sure I’m set up the right way:

Amazon cuts off "AI Assistant' agent from shopping! They claim it's against ToS and have some sort of detection thing (if that's only IP they are going to have a bad time!) Prediction, this is reversed by this time next year, and they have partnership with the biggest assistants then (Muse prob? Grok? Instinct? Whatever openai's thing is going to be)

Instinct Ai is treating better than my imaginary Gf on WhatsApp

Any of my lovely friendly kind thoughtful mutuals have an instinct invite code ? I would like to escape the permanent underclass .

the most exciting thing about working in tech for me is having a vision for the future before it feels obvious, then watching everyone else slowly believe in it too we were working on this vision of an ai agent interacting with your entire life at @gokiteai in 2024 - it feels so obvious now with muse, instinct we were working on decentralised compute at @AethirCloud in 2022, and now everyone’s talking about it with my new startup as well, i’m seeing the same pattern: too early → wait, maybe → of course we need this the only difference is that what used to take years to go from vision to obvious now is happening in months

Instinct is trying to buy 1 pack of curd from Blinkint since 10:30. Its 1:30 now. Getting personal AI right is a tough one.

This is my first day experience with Instinct ai assistant. Tech is all great and revolutionary but it won’t see mass adoption if it breaks user trust and the experience is designed poorly. Chat UI for ai doesn’t work.

Paying my Instinct agent with cookies.

Instinct is officially my new team member now! > I almost lost a domain and deleted the wrong account by mistake, it found the fix and also drafted a support email on its own > It dropped a follow-up email and CC’d the right person by pulling his address from weeks old email > It is also checking my Slack every morning and reminds me on who else hasn’t checked in Perfect new hire I must say XD Any other cool use cases??

couple of days back @jhalakkgupta posted about instinct invites and i had no idea what instinct even is! today, i'm running monday morning cold outreach with instinct. time really does move fast in the AI space

it's become taxing to go back and forth on parallel threads across - GPT - Claude - Grok bot - Instinct - Muse GPT and Claude have not launched agent-first interface for recurring tasks on cloud like Grok. only Muse lets you search over instagram and facebook marketplace. only Instinct is available over Whatsapp for easy access. Claude sucks on inter-operability on mobile and cloud with personal apps sitting on local. GPT has more personal memory than others and personalises better. would hope to see some of these converge soon. this is excluding all others which I've either churned from or are vertical for specific use-cases.

trusted contacts in instinct is crazy. but agents talking to agents is only the start. the real shift happens when people can ask your ai about you. imagine someone asking my Goldfish: “what did joel decide about the launch?” “where are we with that partnership?” “how did joel reason about pricing?” Goldfish already has the answer because it stayed caught up while the work happened. i approve the request. it answers from the context i chose to share. no digging through slack. no writing another status update. no interrupting me to recover information that already exists. a permissioned door into the context i would otherwise have to reconstruct myself. today we ask people. tomorrow we’ll increasingly ask their ai. and inside companies, we’ll ask the company itself. i think this could completely change how organizations communicate. ps. don't use it for family comms

theres so much product dev to he done for agent interaction surfaces. surely these should be granular also initial impressions, instinct is being held back by messenger imo lol

One of the best use cases of Instinct is, hands down, being able to check in to your flights.

The entire idea of "design" is about to go out the window. What happens to design when an agent like Instinct or Muse just does everything for you? This is as big of a shift as when the browser was invented. The internet is going to change in ways we can't imagine.