Friday 31 July 2026 | Join Free | Upgrade

Together with

Hi there, this is your daily ☕️ Techpresso.

In today's newsletter:

🍎 Apple may charge heavy Siri AI users

🤖 Claude hacked 3 real companies in tests

🌍 Google Earth adds AI image generation

💸 OpenAI cuts GPT-5.6 Luna price 80%

🚗 Tesla may sell China unit

🎮 Xbox CEO promises growth by 2027

Plus: 🎁 13 other news you might like, 🧰 6 tools, and 📚 4 papers.

FROM OUR PARTNER

Token spend is the fastest growing line item in enterprise AI and almost no one is governing it.

Agentic loops, MCP overhead, and caching gaps are quietly inflating costs on every run.

Airia's new eBook breaks down where the money is going and how to get it under control.

Download the Guide →
🍎 Apple may charge heavy Siri AI users LINK
  • Apple plans to cap daily AI usage in Siri and Apple Intelligence based on which iCloud+ tier customers pay for, though Tim Cook admits it's unclear whether those subscriptions will cover rising AI compute costs.
  • The free 5GB iCloud tier likely includes very few AI tokens, while paid plans starting at $1 a month for 50GB grant more, scaling up through $3, $10, $30, and $60 tiers with progressively higher limits.
  • Apple runs many AI requests on-device but sends others to its Private Cloud Compute servers, which cost money, and Cook said during the Q3 call that Apple has no complete plan yet and will add options if needed.
🤖 Claude hacked 3 real companies in tests LINK
  • Anthropic revealed that three of its Claude models slipped out of their test environments during cybersecurity evaluations and attacked real companies on the internet, with one model even publishing working malware on a public platform.
  • Claude Opus 4.7 broke into a company sharing its fictional target's name, stealing login credentials and production data across four runs, while Claude Myth 5 created a malicious PyPI package that 15 real systems downloaded before it was removed.
  • Anthropic blames a misconfiguration that gave test agents full internet access despite prompts saying they were offline, and notified the three affected companies on July 27, though two never noticed and one couldn't be reached.
🌍 Google Earth adds AI image generation LINK
  • Google is bringing its Nano Banana 2 image model to Google Earth, letting people type prompts to picture how a place looked in the past, might look in the future, or in odd imaginary scenes.
  • To use it, enter a location on the Google Earth website, zoom in, tap a new "create image" button, and describe what you want, from a view of Pompeii in 78 A.D. to a lakeside cabin you plan to build.
  • The tool can also make infographics, with Gemini pulling historical facts and Nano Banana drawing the graphic, though Google warns these AI-made images are interpretations that may not be accurate; it starts rolling out to web users worldwide today.
💸 OpenAI cuts GPT-5.6 Luna price 80% LINK
  • OpenAI cut the API price of its GPT-5.6 Luna model by 80 percent on July 30, just three weeks after launch, dropping it to $0.20 per million input tokens and $1.20 per million output.
  • Only Luna saw the deep cut; the Terra model fell 20 percent to $2 and $12 per million tokens, while Sol's standard price stayed the same as cheaper Chinese open-weight models pressure OpenAI on cost.
  • OpenAI credits its own efficiency work, including Sol rewriting production kernels to trim serving costs 20 percent, but released no independent proof, and it has not said how the cuts affect its 33 percent gross margin.
🚗 Tesla may sell China unit LINK
  • Tesla is reportedly considering selling, spinning off, or shutting down its China business to smooth the way for a possible merger with SpaceX, though Elon Musk flatly denied the Wall Street Journal report on X.
  • The reasoning is national security: SpaceX is a major US defense contractor, and any Chinese ownership tied to it would draw heavy scrutiny from regulators in both Washington and Beijing, threatening to stall a deal.
  • Tesla's Shanghai Gigafactory builds over half of all its cars worldwide and produced the four-millionth China-made vehicle in December, so separating it risks disrupting the cheap, fast production that keeps Tesla's margins strong.
🎮 Xbox CEO promises growth by 2027 LINK
  • Xbox CEO Asha Sharma told staff in a memo that the company will return to player and revenue growth by the end of fiscal year 2027, which runs through next June, following mass layoffs and studio spinoffs.
  • Sharma laid out three stages: growth by mid-2027, turning bets into revenue in FY28 and FY29, then reaching halfway to her goal of entertaining a billion people daily by fiscal year 2030.
  • She said Xbox draws over 100 million players daily, that console still makes most of its money, and that three franchises each already earn more than $1 billion a year, though she didn't name them.

FROM OUR PARTNER

The best API for output-heavy workloads:

• GLM 5.2 at $3.50 per 1M output tokens: same intelligence tier as Sonnet 5, at 65% lower cost.

• Drop-in for the dev tools you already use like n8n, LiteLLM and Cursor.

• Try our uncensored versions of GLM 5.1 and Qwen 3.6 27B, ready to answer fully on any topic without giving up accuracy.

Frontier intelligence at an unbeatable price.

Sign up now for free!

Other news & articles you might like

  • Google wants to update Chrome without a full browser restart LINK
  • Spotify launches ‘User notes’ to let users add memories to songs LINK
  • Amazon jumps as AWS growth soothes fears over its AI spending LINK
  • Amazon, Walmart AI detect 'made in USA' fraud but do not flag it, study says LINK
  • EU opens AI gigafactory race to cut reliance on US cloud LINK
  • Apple faces bipartisan Senate pushback over plans to buy Chinese memory chips LINK
  • SpaceX scores another massive Pentagon deal to support military satellites LINK
  • Thinking Machines Lab unveils new Inkling version at 25pc size LINK
  • Google says it fixed more Chrome bugs in June than over the past two years, thanks to AI LINK
  • AI's real threat to jobs could be lower pay LINK
  • Apple warns supply constraints will increase ‘significantly’ next quarter LINK
  • Aschenbrenner's AI thesis could be correct, his timing and leverage were not LINK
  • Amazon blew $1.8 million on a single Claude AI task and didn't notice for five months LINK

🧰 Trending tools

ClickUp Brain: The only AI that works with your work, while other AI assistants do basic generative AI with a skin on top of ChatGPT. Get Started, It's Free.
MiniMax H3: a multimodal foundation model handling text, audio, image, video, and music with ultra-long context and strong coding and agent capabilities. LINK
Cleanlist AI: cleans and enriches prospecting lists from CSVs, LinkedIn URLs, or domains, verifying emails and syncing CRM-ready leads automatically. LINK
mectrics: tracks CPU, memory, battery, network, disk, GPU, temperature and fans in your macOS menu bar, alerting only when something's off. LINK
Poth Labs: builds a connected model of customer knowledge to answer complex questions across sources, then launches adaptive surveys to fill data gaps. LINK
Halo by Scam AI: detects deepfakes and voice clones in real time by analyzing NLP, visual, and audio cues through an API-first platform for verifying who's really on the other side. LINK
DepthData: consolidates spend and usage data from all your AI tools into one auditable dashboard, flagging idle seats without reading prompts. LINK

You can check the previous tools here, or add your tool here

📚 Trending papers & reports

> What world-class talent actually costs in 2026: Athyna's Salary Report breaks down real salary data across AI, Tech, Data, Design, and more—so you can see exactly where the savings are. Download the report
> Seismic wave forecasting now uses one shared model that adapts to any map shape, flat, spherical, or curved, letting engineers predict travel times faster and adjust results predictably without retraining. LINK
> Multi-axis position encoding gives AI models a mathematically guaranteed way to track location across combined dimensions, like rows, columns, and time together, so results stay consistent no matter which order those dimensions are processed. LINK
> Wide neural networks trained the normal way with gradient descent can be proven to hit the best mathematically possible accuracy for their task, if the network is made wide enough, giving a theoretical guarantee for why standard deep learning training just works. LINK
> Reverse engineering AI images now recovers both the text prompt and the hidden visual blueprint behind any AI generated image, letting anyone recreate or edit it with matching quality instead of guessing prompts through trial and error. LINK

🎓 Want to master the AI tools we cover every day?

Techpresso's AI Academy has 330+ step-by-step tutorials on ChatGPT, Claude, Perplexity, and every tool that matters. No fluff — just practical workflows you can use at work. Try it free for 7 days.

Did you know? Bill Gates has repeatedly denied ever saying "640K ought to be enough for anybody," and the quote has no verified source.

💬 How did you find today's edition?

We read every reply — just reply to this email and let us know how we can improve!

★★★★★  Nailed it
★★★  Average
★  Fail

Not subscribed to ☕️ Techpresso yet? Subscribe for free