'The Claude Pro Is Consumed Within an Hour': A Week of Coding-Tool Defections Vanishing limits, downgraded models that lecture you while burning your quota, and a memory-hungry desktop app — the developers who live in these tools spent the week doing the maths and switching. #Voices #Claude

'The Claude Pro Is Consumed Within an Hour': A Week of Coding-Tool Defections

Vanishing limits, downgraded models that lecture you while burning your quota, and a memory-hungry desktop app — the developers who live in these tools spent the week doing the maths and switching.

theaidownside.com

Can You Get Your Data Out of an AI Tool? The Right Exists on Paper, the Button Usually Doesn't Data-protection law gives you a right to portability. It was written before chatbots, and the gap it leaves — inferred and generated data — is exactly where your AI history lives. #AIRegulation #Privacy

Can You Get Your Data Out of an AI Tool? The Right Exists on Paper, the Button Usually Doesn't

Data-protection law gives you a right to portability. It was written before chatbots, and the gap it leaves — inferred and generated data — is exactly where your AI history lives.

theaidownside.com

Microsoft Strips Copilot's Free Features — and Puts Deep Research Behind a Subscription Merging two confusing apps is fair enough. Quietly moving a capable free research tool behind a $19.99 plan, and giving podcasts a one-by-one save-or-lose deadline, is the part worth reading…

Microsoft Strips Copilot's Free Features — and Puts Deep Research Behind a Subscription

Merging two confusing apps is fair enough. Quietly moving a capable free research tool behind a $19.99 plan, and giving podcasts a one-by-one save-or-lose deadline, is the part worth reading slowly.

theaidownside.com

'Enshittified at a Surprising Clip': A Week of Hacker News on AI Coding Tools Surprise bills, disappearing quota, mystery-meat UIs and token vomit — the developers who use these tools all day had notes, and they weren't about the robots taking over. #Voices #Cursor

'Enshittified at a Surprising Clip': A Week of Hacker News on AI Coding Tools

Surprise bills, disappearing quota, mystery-meat UIs and token vomit — the developers who use these tools all day had notes, and they weren't about the robots taking over.

theaidownside.com

Why AI Struggles to Guess Your Age — and Why That's a Bias Problem, Not Just an Accuracy One Age-estimation systems are decent in the middle of life and shakiest exactly where the law cares most: the 18 line. Their mistakes don't land evenly on everyone. #Bias #Privacy

Why AI Struggles to Guess Your Age — and Why That's a Bias Problem, Not Just an Accuracy One

Age-estimation systems are decent in the middle of life and shakiest exactly where the law cares most: the 18 line. Their mistakes don't land evenly on everyone.

theaidownside.com

ChatGPT Now Guesses Your Age — and Restricts You by Default if It Thinks You're Under 18 Age prediction reads your topics, your login times and your account age. Get the guess wrong and the way back to the adult product is a selfie or a passport. #Privacy #ChatGPT

ChatGPT Now Guesses Your Age — and Restricts You by Default if It Thinks You're Under 18

Age prediction reads your topics, your login times and your account age. Get the guess wrong and the way back to the adult product is a selfie or a passport.

theaidownside.com

ChatGPT Now Guesses Your Age — and Restricts You by Default if It Thinks You're Under 18 Age prediction reads your topics, your login times and your account age. Get the guess wrong and the way back to the adult product is a selfie or a passport. #Privacy #ChatGPT

ChatGPT Now Guesses Your Age — and Restricts You by Default if It Thinks You're Under 18

Age prediction reads your topics, your login times and your account age. Get the guess wrong and the way back to the adult product is a selfie or a passport.

theaidownside.com

The Upgrade That Wasn’t: When ‘Newer’ AI Feels Like a Downgrade A fortnight of paying users saying the same thing three different ways: the new model is worse, the app quietly removed what they used, and they can’t even tell what they’re running. #Voices #Claude

The Upgrade That Wasn’t: When ‘Newer’ AI Feels Like a Downgrade

A fortnight of paying users saying the same thing three different ways: the new model is worse, the app quietly removed what they used, and they can’t even tell what they’re running.

theaidownside.com

The Right to Be Forgotten Is Hard for AI: Why Deleting Your Data From a Model Isn’t a Delete Button Privacy law assumes data lives in rows you can erase. A neural network turns your data into weights — and getting it back out is a research problem, not a setting. #Privacy #AIRegulation

The Right to Be Forgotten Is Hard for AI: Why Deleting Your Data From a Model Isn’t a Delete Button

Privacy law assumes data lives in rows you can erase. A neural network turns your data into weights — and getting it back out is a research problem, not a setting.

theaidownside.com

Atlassian Now Trains Its AI on Your Work by Default — and Full Opt-Out Is an Enterprise Feature The default is on, the metadata switch is greyed out unless you pay for Enterprise, and the whole thing turns privacy into a plan tier. #Privacy #Atlassian

Atlassian Now Trains Its AI on Your Work by Default — and Full Opt-Out Is an Enterprise Feature

The default is on, the metadata switch is greyed out unless you pay for Enterprise, and the whole thing turns privacy into a plan tier.

theaidownside.com

I just published OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8 medium.com/@theaidownsi...

OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8

Hit your weekly ChatGPT cap and a prompt now offers to restore it to full for about $8. OpenAI never announced it; a user found it at the…

medium.com

The Subscription Squeeze: A Fortnight of Paying-User Gripes Claude users brace for a limit that lapses on the 19th, Perplexity subscribers watch a year-old plan curdle, and everyone does the same grim arithmetic: less for the same money, across three companies at once. #Voices #RateLimits

The Subscription Squeeze: A Fortnight of Paying-User Gripes

Claude users brace for a limit that lapses on the 19th, Perplexity subscribers watch a year-old plan curdle, and everyone does the same grim arithmetic: less for the same money, across three companies at once.

theaidownside.com

How AI Models Can Leak the Data They Were Trained On Large models do not just learn patterns — they memorise fragments of their training data, and researchers keep finding ways to pull those fragments back out. Here is how that works, and why it matters for your data. #Cybersecurity #Privacy

How AI Models Can Leak the Data They Were Trained On

Large models do not just learn patterns — they memorise fragments of their training data, and researchers keep finding ways to pull those fragments back out. Here is how that works, and why it matters for your data.

theaidownside.com

OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8 Hit your weekly ChatGPT cap and a prompt now offers to restore it to full for about $8. OpenAI never announced it; a user found it at the moment they were locked out. #Pricing #OpenAI

OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8

Hit your weekly ChatGPT cap and a prompt now offers to restore it to full for about $8. OpenAI never announced it; a user found it at the moment they were locked out.

theaidownside.com

OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8 Hit your weekly ChatGPT cap and a prompt now offers to restore it to full for about $8. OpenAI never announced it; a user found it at the moment they were locked out. #Pricing #OpenAI

OpenAI Is Testing a Button to Reset ChatGPT’s Limits — For $8

Hit your weekly ChatGPT cap and a prompt now offers to restore it to full for about $8. OpenAI never announced it; a user found it at the moment they were locked out.

theaidownside.com

The Mark and the Meter: A Fortnight of AI Gripes Codex users watch their $200 plans drain in days while Claude starts watermarking everything it touches. Two companies, two kinds of quiet change, one feeling: the product shifted underneath. #Voices #RateLimits

The Mark and the Meter: A Fortnight of AI Gripes

Codex users watch their $200 plans drain in days while Claude starts watermarking everything it touches. Two companies, two kinds of quiet change, one feeling: the product shifted underneath.

theaidownside.com

Your AI-Generated Code Might Not Be Yours The US Copyright Office says purely AI-generated material is not copyrightable. If you use Copilot, Claude or Cursor to write your code, the legal status of what you ship is less clear than your licence agreement suggests. #Copyright #AICoding

Your AI-Generated Code Might Not Be Yours

The US Copyright Office says purely AI-generated material is not copyrightable. If you use Copilot, Claude or Cursor to write your code, the legal status of what you ship is less clear than your licence agreement suggests.

theaidownside.com

Claude’s Invisible Watermark Marks Even Your Own Writing Anthropic’s new EU-compliance watermark stamps every word Claude touches — including text you wrote yourself and asked it to proofread. #Privacy #Anthropic

Claude’s Invisible Watermark Marks Even Your Own Writing

Anthropic’s new EU-compliance watermark stamps every word Claude touches — including text you wrote yourself and asked it to proofread.

theaidownside.com

The Meter, the Limit and the Refusal: A Fortnight of AI Gripes Paying users on four platforms describe the same three feelings this fortnight: the quota shrank, the bill got harder to read, and the tool said no. #Voices #RateLimits

The Meter, the Limit and the Refusal: A Fortnight of AI Gripes

Paying users on four platforms describe the same three feelings this fortnight: the quota shrank, the bill got harder to read, and the tool said no.

theaidownside.com

What ‘AI Safety’ Actually Means (and What It Doesn’t) The phrase does the work of at least three different jobs. Companies rely on you not noticing which one they mean. #Safety #AIRegulation

What ‘AI Safety’ Actually Means (and What It Doesn’t)

The phrase does the work of at least three different jobs. Companies rely on you not noticing which one they mean.

theaidownside.com

Twitch Opted Every Streamer Into Training Amazon’s AI Amazon can now train generative AI on your streams, clips and chat unless you find the toggle. Twitch’s product chief explains why it’s opt-out: if it were opt-in, nobody would say yes. #Privacy #Copyright

Twitch Opted Every Streamer Into Training Amazon’s AI

Amazon can now train generative AI on your streams, clips and chat unless you find the toggle. Twitch’s product chief explains why it’s opt-out: if it were opt-in, nobody would say yes.

theaidownside.com

You Paid for the Big Model. They Quietly Served You the Small One. This fortnight’s complaints, in users’ own words: silent downgrades to cheaper models, usage limits that shrank without a memo, and an assistant that argues with the buttons on your own phone. #Voices #Pricing

You Paid for the Big Model. They Quietly Served You the Small One.

This fortnight’s complaints, in users’ own words: silent downgrades to cheaper models, usage limits that shrank without a memo, and an assistant that argues with the buttons on your own phone.

theaidownside.com

What AI Regulation Actually Protects You From (And What Just Got Delayed) The EU wrote the world’s most ambitious AI law. Its biggest protections for ordinary users were meant to arrive this month — and were quietly pushed to 2027. Here is the honest map. #AIRegulation #Safety

What AI Regulation Actually Protects You From (And What Just Got Delayed)

The EU wrote the world’s most ambitious AI law. Its biggest protections for ordinary users were meant to arrive this month — and were quietly pushed to 2027. Here is the honest map.

theaidownside.com

DeepSeek Turns Its Famously Cheap Tokens Into Peak-Hour Pricing The lab that undercut everyone is raising V4 prices up to fourfold and calling half of it a discount. The timing, just before a reported IPO, is hard to miss. #Pricing #DeepSeek

DeepSeek Turns Its Famously Cheap Tokens Into Peak-Hour Pricing

The lab that undercut everyone is raising V4 prices up to fourfold and calling half of it a discount. The timing, just before a reported IPO, is hard to miss.

theaidownside.com

Refusals, Flattery and the Bill: A Fortnight of AI Gripes This edition’s complaints, in users’ own words: tools that won’t do the ordinary thing, tools that tell you you’re a genius, and tools that quietly cost more than they let on. #Voices #Censorship

Refusals, Flattery and the Bill: A Fortnight of AI Gripes

This edition’s complaints, in users’ own words: tools that won’t do the ordinary thing, tools that tell you you’re a genius, and tools that quietly cost more than they let on.

theaidownside.com

Slopsquatting: When Your AI Assistant Invents a Package and an Attacker Registers It The model hallucinates a dependency, someone claims the name, and your install command runs their code. A supply-chain attack that exists only because AI is confidently wrong. #Cybersecurity #AICoding

Slopsquatting: When Your AI Assistant Invents a Package and an Attacker Registers It

The model hallucinates a dependency, someone claims the name, and your install command runs their code. A supply-chain attack that exists only because AI is confidently wrong.

theaidownside.com

ChatGPT Now Refuses to Write in a Named Author’s Style The behaviour showed up in no changelog — and the copyright timing is hard to miss. #ChatGPT #OpenAI

ChatGPT Now Refuses to Write in a Named Author’s Style

The behaviour showed up in no changelog — and the copyright timing is hard to miss.

theaidownside.com

‘I Didn’t Ask for This’: Users on Being Force-Fed AI This fortnight’s gripes, in users’ own words: no off switch, an opt-out that keeps moving, and the growing suspicion that ‘on by default’ is the whole business model. #Voices #AISearch

‘I Didn’t Ask for This’: Users on Being Force-Fed AI

This fortnight’s gripes, in users’ own words: no off switch, an opt-out that keeps moving, and the growing suspicion that ‘on by default’ is the whole business model.

theaidownside.com

Who Owns the Words That Trained Your AI? The copyright fight over training data isn’t a sideshow to the AI boom — it’s the bill arriving for it. Here is what the courts have actually decided, and what they pointedly haven’t. #Copyright #TrainingData

Who Owns the Words That Trained Your AI?

The copyright fight over training data isn’t a sideshow to the AI boom — it’s the bill arriving for it. Here is what the courts have actually decided, and what they pointedly haven’t.

theaidownside.com

Google Is Testing a Homepage That Hides the Search Button For a slice of signed-out users, the most familiar button on the internet has been swapped for three AI shortcuts. It is only a test — which is exactly why it is worth watching. #AISearch #Google

Google Is Testing a Homepage That Hides the Search Button

For a slice of signed-out users, the most familiar button on the internet has been swapped for three AI shortcuts. It is only a test — which is exactly why it is worth watching.

theaidownside.com