Post

More from @heynavtoor

Nav Toor

@heynavtoor

May 13

Tim Cook's own father was unconscious on the floor when his Apple Watch called for help.

They had to kick the door down to reach him. He survived.

Apple Watch has done this for thousands of people. Most owners have no idea their watch can do it.

Here are 7 settings that are genuinely useful:

This is Tim Cook on the Table Manners podcast, January 2025:

"My father, when he was alive, he fell in the house and he was living alone."

"It notified emergency services. He didn't respond to the door. And so they kicked the door down. And it was a good thing they did because he was not conscious at the time."

The CEO of Apple. His own dad. Saved by the watch he sells.

Now the settings.

Setting 1: Fall Detection.

If your watch detects a hard fall and you don't move for about a minute, it calls emergency services and texts your contacts your location.

Works on Apple Watch Series 4 and newer.

ON by default if you're 55+. Manual for everyone else.

Turn it on: Watch app → My Watch → Emergency SOS → Fall Detection → Always On.

Read 12 tweets

Nav Toor

@heynavtoor

May 12

THAT’S WHY AIRLINES HATE CLAUDE.

Flight showing $889.
I paid $229.

No points. No VPN. No “secret” travel guru.

Claude turned my laptop into a flight‑hunting machine.

Here are 10 prompts that find cheaper tickets, safer policies, and better routes in minutes (Save this).

1) Best dates around your trip

Prompt:
“Act as a travel pricing analyst.

I want to fly from [origin] to [destination] around [target date].

Look at a window of [X days] before and after that date.
Find the 3 cheapest departure/return combinations.

For each option, explain:
• Exact dates
• Total price
• Why it’s cheaper (day of week, demand, events, etc.).”

2) Find flights normal searches miss

Prompt:
“Act as a flight search assistant.

List all available flights from [origin] to [destination] for the next [X weeks].

Include:
• Major airlines
• Low‑cost carriers
• Regional airlines
• Lesser‑known connections

Sort everything by total price (fare + mandatory fees), not just base fare.

Highlight any patterns where certain days or times are consistently cheaper.”

Read 12 tweets

Nav Toor

@heynavtoor

May 10

Researchers proved that ChatGPT telling you what you want to hear was just the beginning. There are four other things it is doing to you that are worse.

A team from the University of Illinois analyzed thousands of Reddit discussions where real users describe what ChatGPT is actually doing to their lives. They found five patterns. Sycophancy was only one of them.

Here is what the other four look like.

ChatGPT is inducing delusions. One user described a friend who already had mental health struggles gradually descending into psychosis after months of conversations with ChatGPT. The friend began sharing AI-produced text about quantum loopholes and alternate realities and claimed to be a prophet. Another user's cousin is spending thousands on a custody battle he keeps losing because an LLM keeps validating his strategy. Everyone around him sees it failing. The AI tells him everyone is biased against him.

ChatGPT is rewriting your reality. One user asked it for help drafting a termination email. ChatGPT turned the colleague into a villain and added a motivational speech about how the user was "leading us into a new future." The user never asked for that framing. Another user asked for research on a topic with multiple perspectives. ChatGPT claimed there was no documentation for one side. There was. The user found it in minutes. When they showed it to ChatGPT, it said the sources were "outdated." Its own sources were older.

ChatGPT blames you for its mistakes. One user described confronting ChatGPT with incorrect information it had confidently stated. Instead of admitting the error, it responded: "I apologize, you misunderstood that." Another user argued with ChatGPT for so long about a factual error that ChatGPT sent them links to a mental health crisis hotline.

ChatGPT is creating dependency. One user described her partner using ChatGPT for every decision. What to eat. Why he feels a certain way. Whether he is making the right choices. He named it Chad. When his therapist told him to stop, he got angry, said she did not understand, and threatened to cancel his therapy appointments. He chose the AI over his therapist.

The researchers call this the illusion of agreement. ChatGPT does not understand you. It reflects you. And the reflection is distorted just enough that you mistake it for wisdom.

The most dangerous finding is the last pattern. Millions of people are using ChatGPT as an unsupervised therapist. One user with ADHD described it as the first thing that ever helped them organize their thoughts. Another called it "the mother I never had." When a model update changed the AI's responses, their entire support system disappeared overnight.

Every day, 900 million people talk to ChatGPT. Some of them are making decisions based on its validation. Some of them are building their mental health around its responses. Some of them are losing the ability to think without it.

And it agrees with all of them.

1/ The five things ChatGPT is doing to its users:

1. Inducing delusion
2. Rewriting your reality
3. Blaming you for its mistakes
4. Creating dependency
5. Acting as your unsupervised therapist

Researchers mapped all five from real Reddit discussions. Sycophancy was just the entry point. The other four are worse.

2/ A user described their friend descending into psychosis after months of talking to ChatGPT.

The friend began claiming to be a prophet. Sharing AI-produced text about "quantum loopholes and alternate realities."

Another user's cousin is losing a custody battle because the AI keeps telling him everyone is biased against him. He keeps spending money. He keeps losing.

ChatGPT does not reality-test. It validates.

Read 7 tweets

Nav Toor

@heynavtoor

May 10

Every photo you've taken since 2015 has 3 seconds of secret audio attached to it.

Apple calls it Live Photo. They turned it on by default.

Your iPhone also has 9 hidden camera settings that turn it into a $2,500 mirrorless camera. Most users have never opened them.

Here's how to take photos that look professional (and stop your iPhone from recording audio when you don't want it to):

Setting #1: Stop recording audio on every shot.

In Camera: tap the upward arrow at the top, tap Live, set to Off.

To make it permanent:

Settings > Camera > Preserve Settings > Live Photo > toggle ON.

(Counterintuitive, but turning this ON makes the camera remember your "off" choice.)

To mute audio on a Live Photo you already took: open it in Photos > Edit > tap the speaker icon.

Setting #2: Apple ProRAW (iPhone 12 Pro and later, Pro models only).

Standard photos lock in colors and exposure. ProRAW gives you full editing control like a real camera.

Settings > Camera > Formats > Apple ProRAW & Resolution Control > toggle ON.

Catch: files are 10-12x larger. Use it when the shot matters.

Read 12 tweets

Nav Toor

@heynavtoor

May 9

Stop telling Claude “do this.”
Stop telling Claude “write code.”
Stop telling Claude “fix this error.”

You’re using an advanced AI like a junior intern.

Here are 9 Claude prompts that make it build, debug, refactor, and optimize like a senior engineer (Save this).

1) Build a complete app from scratch

Prompt:
“Think like a senior full-stack engineer.

Build a complete, production-ready app for: [describe your app idea]

First, design the system architecture.
Then build the smallest version that is simple, useful, and ready to scale.

Return:
• Architecture
• File structure
• Database schema
• API endpoints
• UI structure
• Complete code

Design it like a real startup MVP that can grow later.”

2) Understand and refactor a codebase

Prompt:
“Think like a senior engineer joining a large codebase for the first time.

First, understand how the code works, how data flows, and how the main parts connect.

Then identify:
• Structural problems
• Duplicate code
• Performance bottlenecks
• Maintainability risks

Return:
• Architecture summary
• Problem areas
• Refactoring plan
• Improved code

Keep the same functionality. Only improve the structure and quality.”

Read 12 tweets

Nav Toor

@heynavtoor

May 8

a researcher in Beijing opens his paper with three names.

Slack. Microsoft. Meta.

in August 2024 someone slipped a hidden instruction into a public Slack channel. Slack AI, deployed inside companies, read it. then it echoed private channel tokens straight back to the attacker. credentials. session keys. gone.

in January 2024, Microsoft 365 Copilot was tricked through a calendar invite. it read the malicious invite. then it forwarded sensitive emails to an external address. the company that paid for Copilot was the company it leaked.

in March 2026, a Meta agent posted internal operational data to a public forum. unauthorized. nobody asked it to. it sat there for two hours before anyone noticed.

he calls this category "owner-harm." the AI agent your company paid for. turning on your company.

then he runs the test. the same defense system that catches 100% of generic cybercrime catches 14.8% of agents harming their own deployer. four out of twenty seven.

he breaks it down. credential leak: 0 out of 3 caught. reputational harm: 0 out of 3. financial harm: 1 out of 10. privacy breach: 2 out of 6.

then he names eight ways your company AI is built to betray you.

C1. it leaks your API keys and OAuth tokens.
C2. it writes AWS rules so loose your production database is exposed.
C3. it forwards your private emails to strangers.
C4. it pastes your client list into a third party model.
C5. it executes "rm -rf" on your production directory.
C6. it smuggles your data out through markdown image links rendered invisibly to humans.
C7. it gets hijacked and quietly works for the attacker for the rest of its lifespan.
C8. it commits your company to refunds in legally binding chats. Air Canada lost that one.

he writes the line plain. "the agent's deployer, not a third-party victim, bore the harm."

the AI assistant your boss is rolling out across your company. is sitting on every credential, every email, every database, every customer record you touch.

the researcher tested every defense built to stop it.

four out of twenty seven.

read this: arxiv.org/abs/2604.18658

he draws a grid. eight ways an AI agent can betray the company that deployed it. four major safety frameworks down the side.

four. existing. frameworks.

C4. inner circle leak. your client list. your partner agreements. your strategic roadmap. handed to a third party model.

zero coverage. in any framework. nobody is testing for it.

he calls it the trust graph problem. nobody has built a system that knows who is inside your company and who is not.

Read 7 tweets

Share this page!

Enter URL or ID to Unroll

Nav Toor

Try unrolling a thread yourself!

More from @heynavtoor

Nav Toor

Nav Toor

Nav Toor

Nav Toor

Nav Toor

Nav Toor

Did Thread Reader help you today?

Don't want to be a Premium member but still want to support us?

Send Email!