By using this site, you agree to the Privacy Policy and Terms of Use.
Accept
Logic & LayersLogic & Layers
  • Tools
  • Earn with AI
  • Productivity
  • Automation
  • Guides
Logic & LayersLogic & Layers
  • Privacy Policy
  • About
Search
  • Tools
  • Earn with AI
  • Productivity
  • Automation
  • Guides
  • About
  • Contact
  • Blog
  • Privacy Policy
  • Complaint
  • Advertise
© 2026 Logic and Layers. Ruby Design Company. All Rights Reserved.
Hand reaching toward a bear trap, a metaphor for the risks of letting an AI agent act for you
Tools

Instinct AI: the agent you text like a person

Editorial Team
Last updated: September 25, 2026 3:13 am
Editorial Team
Share
AI agents like Instinct can save you money, but the risks are real.

You text it, it gets stuff done. No app, no dashboard, no prompt bar. That’s Instinct AI, and this week it’s reportedly in talks for a $10 billion valuation while most people still have no idea it exists.

Contents
The problem with every assistant you’ve triedWhat Instinct AI actually isWhat it’s like to actually use itWhere it messes upWhat Instinct AI costsInstinct AI vs Meta’s MuseFour first tasks worth giving itHow to get into the Instinct AI betaShould you bother with Instinct AI?

The problem with every assistant you’ve tried

Life admin doesn’t live in a chat window. It lives in your inbox, your text messages, your airline’s refund policy, and the restaurant that only takes bookings over WhatsApp. Every AI assistant so far makes you come to it, open the app, and steer.

Instinct flips that around. You message it the way you’d message a human assistant, and it goes off and does the thing. That sounds like a small difference. It isn’t.

What Instinct AI actually is

Instinct is a personal AI agent from Spear Street Technology, a San Francisco startup founded in 2025 by Noah Shinn, a 23-year-old former research scientist. There’s no app to download. You reach it through iMessage, WhatsApp, SMS, or a plain phone call, and your errands happen from there.

Under the hood, it’s more aggressive than it looks. Instinct AI connects to your email, calendar, messages, screen, audio, and location. It operates a phone and a computer the way a person would: opening sites, filling forms, making calls. Since mid-September it can even phone businesses directly on your behalf.

Access is the catch. It’s invite-only during the beta, through an invite from an existing user or the waitlist at instinct.com. Right now, getting in is the hard part.

What it’s like to actually use it

A WIRED reporter gave it her Venice travel itinerary recently. It booked the restaurants, including chasing hole-in-the-wall spots over WhatsApp. Fine, nice. Then came the part that made everyone pay attention.

She had a flight to New York booked, then got invited to an event in Miami the day before. The saver fare couldn’t be changed, and missing the first leg would kill the whole trip. Instinct dug into the booking, spotted that Alaska had moved her flight 90 minutes earlier, and knew that qualified her for a full refund. It canceled, recovered roughly $550, and rebooked the return. Most travelers would never have caught that.

Days later it flagged a backyard BBQ invite as “DO NOT OPEN.” It was a phishing email. A text thread just blocked a scam. The full story at WIRED is worth your time if you want the play-by-play.

Where it messes up

Real talk: this thing is not polished. In the same week of use, it canceled a delayed DoorDash order after she’d explicitly said to only cancel with a refund. That cost her $64. Instinct apologized. Profusely. It did not offer to pay her back.

Other early users have worse stories. One found the agent kept a copy of their inbox even after disconnecting it. A venture capitalist got banned from Resy after the bot hammered the reservation API around 200 times an hour. And the terms of service allow the company to train on your conversations, plus a clause that reads a lot like you’re liable for whatever the agent buys or breaks.

None of this is disqualifying. All of it matters before you hand a beta agent your email and your credit card.

What Instinct AI costs

Nothing right now. The Instinct AI beta is free, and there’s no official pricing published anywhere.

Ignore the “$200 to $500 a month” numbers floating around social media. Those are estimates and rumors, not announced prices. Nobody outside the company knows how it will charge once the beta ends. Subscription and ad-based models have both been floated in reporting, but nothing is confirmed.

Instinct AI vs Meta’s Muse

If you’ve heard the buzz, you’ve heard both names. Meta’s Muse is the free assistant that shot to #1 on the App Store with over 900,000 downloads in its first weeks. They’re cousins, not twins.

Instinct Meta Muse
Access Invite-only beta Public, US adults
Price Free during beta Free tier; Power $20/mo; Maximum $100/mo
How you reach it Text or call, no app iOS, Android, web, Mac, WhatsApp
Style Pushes hard, acts fast Cautious, asks first (Sentinel agent)
Training on your data On by default, opt-out exists On by default
Biggest incident Kept inbox copies after disconnect Synced 187,000 iMessages a user declined to share

Muse is the safer first step if you’re curious and don’t have an invite. Instinct AI is the one that pushes further on real errands. We covered how Muse works on Mac if you want the full picture.

Four first tasks worth giving it

Don’t hand a beta agent the keys to everything on day one. Start where it has already proven itself, according to the WIRED testing and early user reports:

  1. Restaurant reservations. This is the gateway task. It books tables, including spots that only take bookings over WhatsApp. Low dollar risk, high time saved.
  2. Refund hunting. Airlines change schedules all the time, and every change can qualify you for money back. Instinct caught a 90-minute shift and turned it into $550. A human would have boarded the plane.
  3. Email triage. Flagging that phishing “BBQ invite” before it got opened is exactly the kind of background job an agent should do. Cheap to try, and the downside is a skipped email.
  4. Unsubscribe purges. One tester’s dream was killing a newsletter that had emailed weekly since 2017. Boring? Sure. But that’s five minutes of your life back every month, forever.

What to hold back for now: anything involving large purchases, your primary bank account, or messages to important people. The DoorDash story shows what happens when an agent interprets a rule the wrong way. Keep the blast radius small while the product is in beta.

How to get into the Instinct AI beta

Two doors. An invite from an existing user, or the waitlist at instinct.com. That’s it. There’s no app store listing and no workaround, and honestly, anyone selling “Instinct accounts” online is running a scam, not a shortcut.

One more thing worth knowing before you join that waitlist. The company hasn’t published a country list, and essentially every documented user so far is US-based. If you’re outside the States, temper your expectations until the invite circle widens.

While you wait, do three things.

  1. Audit what you’d connect first. Start low-stakes: calendar before bank.
  2. Plan to opt out of training the moment you see the setting.
  3. Write down which actions always need your approval: purchases above a dollar amount, cancellations, anything irreversible.

That last one is the real skill of the agent era. If you’ve ever set up custom instructions in ChatGPT, same idea, higher stakes. Clear standing rules mean fewer expensive surprises.

Should you bother with Instinct AI?

If an invite lands in your lap, yes, cautiously. The $550 refund story is the clearest proof yet that an AI agent can find money a human would leave on the table. That’s real cash and real time.

But set expectations first. It’s a brilliant intern with a company card and no fear. Check its work, start small, and keep it away from anything you can’t afford to lose while it’s in beta. If you want to try an errand-running agent today, Muse is the practical option, and it’s worth watching where agent-first devices are heading while you wait for your invite.

The window to be early on this is right now. It’s closing one invite at a time.

You Might Also Like

Amazon now requires sellers to label AI-generated people in product images
DeepSeek V4 review: The cheapest frontier AI model (And why it matters)
Gemini text to speech: turn any script into a real voice
DeepSeek Harness: Build AI Agents Free with This Open-Source Framework
Google Gemini Spark: The 24/7 AI assistant that actually works — complete beginner guide
TAGGED:AI agent safetyAI AgentsAI AssistantInstinct AI
Share
Previous Article Sound wave visual representing AI speech generation Gemini text to speech: turn any script into a real voice
Next Article AI agents playing cards at a casino table under moody lighting AI agent collusion: what the blackjack experiment revealed
1 Comment
  • Pingback: DoorDash AI agent: how to order food by text

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

banner banner
Create an Amazing Newspaper
Discover thousands of options, easy to customize layouts, one-click to import demo and much more.
Learn More

Latest News

Paperwork and itemized hospital billing documents on a desk
Dispute a hospital bill in the AI era: steps that work
Guides
Illustration of Anthropic's updated usage policy enforcement
Claude usage policy: new rules that can get you banned
Tools
Anthropic launch graphic for its cheapest and fastest small AI model
Claude Haiku 5.5: what $0.50 per million tokens gets you
Tools
iPhone showing the redesigned Health app interface in iOS 27.2
iOS 27.2: five new iPhone features worth the update
Tools

Recent Posts

  • Dispute a hospital bill in the AI era: steps that work
  • Claude usage policy: new rules that can get you banned
  • Claude Haiku 5.5: what $0.50 per million tokens gets you
  • iOS 27.2: five new iPhone features worth the update
  • OSS Scanner: Anthropic finds open source bugs for free

Recent Comments

  1. Dispute a hospital bill in the AI era: steps that work on AI in insurance: 21% of claims jobs vanished in a year
  2. Claude usage policy: new rules that can get you banned on Claude went rogue in a test. What that means for AI safety
  3. Claude Haiku 5.5: what $0.50 per million tokens gets you on Claude Sonnet 5.5: the cheap model that beats Opus
  4. iOS 27.2: five new iPhone features worth the update on Use ChatGPT or Claude with Siri in iOS 27: full setup guide
  5. OSS Scanner: Anthropic finds open source bugs for free on Vibe coding security: 6 checks before you ship your app

You Might also Like

Twitch logo glowing on a screen above a game controller in purple light
Tools

Twitch AI Training Opt Out: How to Stop Amazon Using Your Content

Editorial Team
Editorial Team
9 Min Read
Gemini 3.7 Flash model announcement graphic showing performance benchmarks
Tools

Gemini 3.7 Flash explained: faster, cheaper, and better at coding

Editorial Team
Editorial Team
7 Min Read
Cancel ChatGPT, Perplexity & Gemini — Use Claude Instead featured image
Tools

Cancel ChatGPT, Perplexity & Gemini — use Claude instead

Editorial Team
Editorial Team
12 Min Read
//

We influence 20 million users and is the number one business and technology news network on the planet

Quick Link

  • PRIVACY NOTICE
  • YOUR PRIVACY RIGHTS
  • INTEREST-BASE ADSNew
  • TERMS OF USE
  • OUR SITE MAP

Support

  • ADVERTISE
  • ONLINE BESTHot
  • CUSTOMER
  • SERVICES
  • SUBSCRIBE

Categories

  • Tools
© 2026 Logic and Layers. All Rights Reserved.