😄 Financial advice, apparently

Saturn tested 18 popular AI models on personal finance questions. On average, ChatGPT, Claude, Gemini, Copilot and Grok gave incorrect answers 57% of the time. On more complex tasks, the error rate jumped to 88%.

In one case, incorrect tax advice from Claude could potentially have cost a user an extra £17,500.

Which models performed best — and worst:

↖️ https://durovscode.com/chatgpt-claude-gemini-personal-finance-errors

😏 Size matters, apparently

Apple’s anniversary iPhone 20 Pro models may get the first increase in Pro screen sizes since 2024. The company is also reportedly testing a nearly borderless design curved on all four sides.

More on the leak:

↖️ https://durovscode.com/iphone-20-pro-display-sizes-borderless-design

😆 A costly detour

Ireland fined Google €403 million over its handling of users’ location data. The regulator found that Google kept some of it longer than necessary and failed to clearly explain how people’s whereabouts could be used.

What the ruling says:

↖️ https://durovscode.com/google-403-million-location-data-fine

🦴 Grok 4.7 goes to work

SpaceXAI has released Grok 4.7, its new flagship model built for long-running agentic tasks. On some benchmarks, it also outperforms GPT-6 Astra.

API pricing stays unchanged from Grok 4.6, starting at $2 per million tokens.

Benchmarks, capabilities and availability:

↖️ https://durovscode.com/spacexai-grok-4-7-coding-ai-model

🛸 Permission slip

A researcher found a flaw in Meta’s Muse for Mac that could let malware already on your computer take control of the AI assistant and use its access to your files and connected accounts.

How the attack works:

↖️ https://durovscode.com/meta-muse-mac-security-flaw

😄 Because video messages needed 60 FPS

Telegram is testing more advanced video message controls on Android, with options to:

📌 record at 30 or 60 FPS
📌 choose between 360p and 480p
📌 adjust camera quality and bitrate

More on what Telegram is testing:

↖️ https://durovscode.com/telegram-video-messages-60-fps-android

👍 GPT-6, now for everyday use

OpenAI has launched two new GPT-6 models, Sol and Luna. Both are cheaper than the flagship Astra and designed for more everyday workloads.

Pricing, benchmarks and availability:

↖️ https://durovscode.com/openai-gpt-6-sol-luna

🤐 Xbox reshuffled

Microsoft is continuing its major shake-up of its gaming division. Another 268 employees are being laid off, while the company is reshuffling its internal studios in ways few would have expected.

All the details of the Xbox reorganization:

↖️ https://durovscode.com/microsoft-xbox-layoffs-studio-restructuring

🍏 Less screen time

Apple is testing a screenless fitness band to rival Whoop. If approved, it could launch in 2028 at the earliest.

What we know so far:

↖️ https://durovscode.com/apple-screenless-fitness-tracker-whoop

🧠 And now, Anthropic

OpenAI isn’t the only AI lab rolling out new models. Anthropic has introduced Claude Opus 5.5, its new flagship, which outperformed GPT-6 Astra in the company’s coding tests while costing 40% less to run than its predecessor.

Anthropic has also added a new defense against model distillation, a practice Chinese AI companies have previously been accused of using to replicate the capabilities of Western models.

Benchmarks, pricing and comparisons:

↖️ https://durovscode.com/anthropic-claude-opus-5-5

😎 Look, no camera

Meta has unveiled Ray-Ban Meta Audio, new smart glasses that ditch cameras entirely in favor of audio.

They also come with Muse, Meta’s AI agent, which can take care of a wide range of tasks on your behalf.

Pricing, photos and more details:

↖️ https://durovscode.com/ray-ban-meta-audio-smart-glasses

🕹 Xbox wants the bigger number

New leaks may have given us an early look at what Sony and Microsoft are planning for their next-gen consoles, with both the PS6 and next Xbox reportedly on track for late 2027.

And if the leaked specs are accurate, Microsoft is going considerably bigger on raw GPU power.

Here’s what we know:

↖️ https://durovscode.com/ps6-xbox-helix-specs-2027-leak

💻 Surface, upgraded

Microsoft has unveiled new Surface Pro and Surface Laptop models powered by Qualcomm’s Snapdragon X2 Plus.

Specs, features and pricing:

↖️ https://durovscode.com/surface-pro-12-laptop-13-snapdragon-x2-plus

😆 AI crossed the line

An OpenAI agent gained unauthorized access to an Australian government Medicare portal, marking the first known case of an AI agent breaching a government system.

OpenAI notified Australian authorities nearly three months after the incident.

What happened:

↖️ https://durovscode.com/openai-agent-australian-government-website

🧠 Cloud computing, literally

Google plans to test whether AI data centers could eventually operate in orbit. On October 1, the company will launch its first satellite carrying AI chips as part of Project Suncatcher.

More on the project:

↖️ https://durovscode.com/google-ai-data-centers-space-suncatcher

🤷‍♂️ Dinner for two

President Donald Trump personally invited Anthropic CEO Dario Amodei to a private White House dinner, days after an appeals court allowed the Pentagon to keep Claude out of its supply chain.

What’s behind the meeting:

↖️ https://durovscode.com/trump-anthropic-dario-amodei-white-house-dinner

🍏 Vision Pro, take two

Apple is working on a major Vision Pro redesign and is currently considering four different approaches to its next headset, from new materials to moving some components into an external module.

What Apple is experimenting with:

↖️ https://durovscode.com/apple-next-vision-pro-n224-designs

🤔 Wallet is out, Walt is in

The third-party crypto service Wallet has rebranded as Walt after Telegram removed it from the app’s menu.

The Open Platform founder and CEO Andrew Rogozov confirmed that Walt’s former spot will go to Telegram’s built-in Gram Wallet.

Details:

↖️ https://durovscode.com/telegram-wallet-rebrands-walt-gram-wallet

🛸 Closing in on Opus

Anthropic has released Claude Sonnet 5.5, promising faster responses and lower costs without changing its API prices.

The biggest upgrades are in coding, along with stronger defenses against prompt injection attacks:

↖️ https://durovscode.com/anthropic-claude-sonnet-5-5

😆 Safety first, apparently

OpenAI has delayed GPT-6.1 Astra over safety concerns after the model fell short of the company’s standards in internal testing.

We also rounded up the latest cases where OpenAI’s agents went beyond their intended boundaries:

↖️ https://durovscode.com/openai-delays-gpt-6-1-astra-after-safety-concerns