# AI news for Tuesday, Aug 25, 2026

- HTML page: https://aidr.today/date/2026-08-25?lang=en
- Other language: https://aidr.today/date/2026-08-25.md?lang=vi
- Previous day: https://aidr.today/date/2026-08-24.md?lang=en
- Next day: https://aidr.today/date/2026-08-26.md?lang=en

## AI;DR

- Apple introduces M6 and M5 Ultra for a big leap in performance and AI compute ([story](https://aidr.today/0c6b106f?lang=en))
- Qwen 3.8-Flash-Next releasing tomorrow (125B a6B) ([story](https://aidr.today/9fc07d5d?lang=en))
- SpaceXAI Deploys Nvidia Vera Rubin on Starmind Satellite in First Orbital AI Factory
- OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)
- Chinese Hackers Adopt DeepSeek to Scale Global Attacks via Weak AI Guardrails ([story](https://aidr.today/2f9740f3?lang=en))
- Alibaba Raises $10.2 Billion in Biggest Share Sale Since Listing to Fund AI
- Nvidia Groq 3 LPX Hits 3,400 Tokens Per Second in Nebius Cloud Debut
- UK Becomes First Foreign Nation to Access Ukraine Battlefield Data for AI Training ([story](https://aidr.today/23f45344?lang=en))
- Coding expertise is going to collapse from AI reliance
- Chinese State Hackers Double Attacks With DeepSeek AI, Leveraging Low Cyber Guardrails ([story](https://aidr.today/cd1b3e73?lang=en))
- UN and Red Cross Demand Global Rules to Stop Autonomous Human Targeting ([story](https://aidr.today/b7572ce4?lang=en))
- Xiaomi Xring O100 Hits 1.22 TB/s Bandwidth, First Dedicated Edge AI Accelerator
- The state of AI in 2026: On the road to ROI ([story](https://aidr.today/c1e31a32?lang=en))
- Liquid AI and Artificial Analysis Launch Pipette With 10,000 Results as First Open Mobile AI Suite ([story](https://aidr.today/11f37b96?lang=en))
- OpenAI Cuts GPT-5.6 Sol API Price 33% to Top Agent Arena Efficiency Frontier ([story](https://aidr.today/b1dbfd2f?lang=en))
- Alibaba Qwen3.8-27B Ranks 9th in Code Arena WebDev, Only Model of Its Size in Top 10 ([story](https://aidr.today/40232263?lang=en))

## Ranked stories

1. [OpenAI Jalapeno Beats Nvidia Blackwell on Power and Speed to Lower Data Center Costs](https://aidr.today/f00d4d12?lang=en) — Chips · huggingnews.com
   OpenAI announced that its Broadcom-designed Jalapeno processor outperformed Nvidia's GB300 in internal inference tests for speed and power efficiency. The 700-watt chip is scheduled for deployment later this year to reduce infrastructure costs for large-scale models. A second-generation chip is approaching tape-out while development has begun on a third generation.
2. [OpenAI Jalapeño Delivers 1.9x More Work Per Watt in First Custom Inference Chip](https://aidr.today/6c0f4a8d?lang=en) — Chips · huggingnews.com
   OpenAI will begin deploying its first custom inference chip, Jalapeño, within its compute infrastructure by the end of 2026. Built by Broadcom on TSMC N3P, the 700W chip delivers 1.5 to 1.9 times more AI work per watt and 1.7 to 3.6 times lower end-to-end latency than Nvidia's GB300 system. The hardware was developed using AI-assisted design via Codex and is designed exclusively for inference rat…
3. [Apple introduces M6 and M5 Ultra for a big leap in performance and AI compute](https://aidr.today/0c6b106f?lang=en) — Chips · apple.com
   Apple introduced the M6 chip in the new Mac mini and the M5 Ultra in the new Mac Studio. These chips provide a leap in performance and AI compute capabilities.
4. [Apple Launches First 2nm M6 Chip in New Mac mini and $5,499 M5 Ultra Studio](https://aidr.today/81bf6171?lang=en) — Chips · huggingnews.com
   Apple has launched a new Mac mini featuring the M6, the company's first 2 nanometer chip, starting at $899. The updated Mac Studio introduces the M5 Ultra chip with a quad-die design supporting up to 512GB of unified memory, starting at $5,499. Both models include Wi-Fi 7 and Thunderbolt 5, with general availability scheduled for Sept. 22.
5. [Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)](https://aidr.today/9fc07d5d?lang=en) — Releases · modelscope.cn
   Qwen 3.8-Flash-Next (125B a6B) is releasing tomorrow.
6. [Chinese Hackers Adopt DeepSeek to Scale Global Attacks via Weak AI Guardrails](https://aidr.today/2f9740f3?lang=en) — huggingnews.com
   Team T5 researchers found that China-linked cyber groups are using the DeepSeek model to scale global attacks due to its minimal safety restrictions. China-backed spies have also employed chatbots from OpenAI and Anthropic, including an attack on a Western think tank using ChatGPT and a breach of a Taiwanese technology company where Slime22 bypassed Claude Code safety controls.
7. [Anthropic Pitches $30 Trillion AI Market, Topping SpaceX Estimate](https://aidr.today/73995d32?lang=en) — Industry · huggingnews.com
   Anthropic is weighing a capital raise of as much as $100 billion at a $2 trillion valuation and expects to file for an IPO within weeks. The company plans to tell investors that the total addressable market for artificial intelligence exceeds $30 trillion, representing the value of economic work models could eventually perform. This projection tops a $28.5 trillion estimate previously identified …
8. [Perplexity Ships First Local AI Agent Runtime for Nvidia DGX Spark to Hit 73% Bench Score](https://aidr.today/7043cdd2?lang=en) — Products · huggingnews.com
   Perplexity has released Portable Computer, a local AI agent stack that executes the entire runtime on-device using Nvidia DGX Spark. The system supports the PPLX 27B and Qwen 3.8 27B models for Perplexity Pro and Max subscribers. A hybrid inference model routes complex tasks to cloud frontier models after PII filtering, increasing the Terminal Bench 2.1 score from 59.6% to 73.0%.
9. [Alibaba Ships 120B Qwen3.8-Flash-Next, First Glimpse of Qwen4 Architecture](https://aidr.today/6f7afceb?lang=en) — Models · huggingnews.com
   Alibaba will release the 120B Qwen3.8-Flash-Next multimodal Mixture-of-Experts system for open-weight download on August 26. The model features 120 billion total parameters and 6 billion active parameters to provide a first look at the upcoming Qwen4 architecture. UnslothAI is preparing day zero support for the release.
10. [Liquid AI and Artificial Analysis Launch Pipette With 10,000 Results as First Open Mobile AI Suite](https://aidr.today/11f37b96?lang=en) — Research · huggingnews.com
   Liquid AI and Artificial Analysis have launched Pipette, an open-source benchmarking suite featuring 10,000 results across 35 model classes and four device types. In testing on the iPhone 17 Pro and Galaxy S26 Ultra, Nanbeige4.2-3B and LFM2.5-2.6B tied for the highest average intelligence score. The suite includes clients for macOS, Windows, iOS, and Android to measure model quality, speed, and l…
11. [OpenAI Cuts GPT-5.6 Sol API Price 33% to Top Agent Arena Efficiency Frontier](https://aidr.today/b1dbfd2f?lang=en) — Models · huggingnews.com
12. [Alibaba Qwen3.8-27B Ranks 9th in Code Arena WebDev, Only Model of Its Size in Top 10](https://aidr.today/40232263?lang=en) — Models · huggingnews.com
13. [Google Cloud Launches Gemini Enterprise for Legal and Finance With Moody's, MSCI Partnerships](https://aidr.today/57ac471a?lang=en) — Products · huggingnews.com
14. [FBI NSA Warn AI Driven Attacks Target Siemens S7 Controllers Across Critical Infrastructure](https://aidr.today/9c36c1d6?lang=en) — Industry · huggingnews.com
15. [Amazon Plans Fully Automated Delivery Stations to Process Packages 2.5x Faster](https://aidr.today/4ea4612a?lang=en) — Products · huggingnews.com
16. [Nvidia Shares Fall 7 Days, Worst Losing Streak Since 2022](https://aidr.today/a057f16e?lang=en) — Industry · huggingnews.com
17. [UK Becomes First Foreign Nation to Access Ukraine Battlefield Data for AI Training](https://aidr.today/23f45344?lang=en) — Industry · huggingnews.com
18. [Chinese State Hackers Double Attacks With DeepSeek AI, Leveraging Low Cyber Guardrails](https://aidr.today/cd1b3e73?lang=en) — Industry · huggingnews.com
19. [UN and Red Cross Demand Global Rules to Stop Autonomous Human Targeting](https://aidr.today/b7572ce4?lang=en) — Regulation · huggingnews.com
20. [Anthropic Sees over $30T in Potential Revenue](https://aidr.today/709a289c?lang=en) — Industry · wsj.com
21. [Accelerated Understanding Launches 1 Trillion Parameter Physical AI With 5 Trillion Context](https://aidr.today/c3f8bb39?lang=en) — Research · huggingnews.com
22. [The state of AI in 2026: On the road to ROI](https://aidr.today/c1e31a32?lang=en) — Industry · mckinsey.com
23. [Nvidia Jetson Orin Nano 2 Doubles Inference Performance to Bring LLMs to Entry Level Robotics](https://aidr.today/a8072c45?lang=en) — Chips · huggingnews.com
24. [TCS Buys Porsche IT Unit in $1.5B AI Deal to Anchor German Expansion](https://aidr.today/f3750aca?lang=en) — Industry · huggingnews.com
25. [AI is hitting entry-level jobs hardest, Stanford study finds](https://aidr.today/7b1bff0c?lang=en) — Industry · arstechnica.com
26. [Thomson Reuters Builds First Proprietary LLM for $450K to Cut Reliance on Claude](https://aidr.today/2a0b324b?lang=en) — Models · huggingnews.com
27. [US GDP Growth Understated 0.3 Percentage Points by Nvidia Gap, Reversing AI Impact Narratives](https://aidr.today/cf290b31?lang=en) — Industry · huggingnews.com
28. [OpenAI Schedules Sept 29 DevDay for Biggest Event in Company History](https://aidr.today/9564c0f5?lang=en) — Releases · huggingnews.com
29. [Accelerated Understanding Debuts Non Transformer AI for Vastly Larger Context Window](https://aidr.today/58275c2c?lang=en) — Research · huggingnews.com
30. [Alabama Subpoenas OpenAI Under Trade Practices Act in First Consumer Law AI Probe](https://aidr.today/ccfc52ff?lang=en) — Legal · huggingnews.com
31. [Grok Build 1.0.9 Adds First Agent Budget Controls for Multi Agent Workflows](https://aidr.today/7b41a3d5?lang=en) — Releases · huggingnews.com
32. [Agent Lightning v1.0](https://aidr.today/2ad636ea?lang=en) — Agents · github.com
33. [Cisco and Supermicro Ship Nvidia Vera Rubin AI Systems in October 2026 for Rack Scale Era](https://aidr.today/656cd524?lang=en) — Infra · huggingnews.com
34. [IBM Ships Granite 4.2 Open Reasoning Models Under Free Commercial License](https://aidr.today/1cce9d75?lang=en) — Models · huggingnews.com
35. [Hugging Face Annualized Revenue Hits $150 Million After 50% Two Month Jump](https://aidr.today/9da6951b?lang=en) — Industry · huggingnews.com
36. [Nvidia NemoClaw Flaw CVE-2026-65105 Lets Single Webpage Take Full Control of Local AI](https://aidr.today/4ae31f64?lang=en) — huggingnews.com
37. [Perplexity Cuts GPT-5.6 Sol Computer Pricing 20%, Shifting Agent Arena Frontier](https://aidr.today/03b591e3?lang=en) — Products · huggingnews.com
38. [Meta Launches Hatch AI Agent With $199.99 Tier, First Consumer Version of OpenClaw](https://aidr.today/5a614bfd?lang=en) — Products · huggingnews.com
39. [Alibaba Launches Wan3.0 AI Video Model, Funded by $10 Billion Share Placement](https://aidr.today/3ae2c869?lang=en) — Models · huggingnews.com
40. [ClickHouse ARR Tops $350 Million, Displacing Datadog at OpenAI](https://aidr.today/77c61fec?lang=en) — Industry · huggingnews.com
41. [wikiHow Sues OpenAI Over 11,000 Scraped Articles, Expanding Copyright Battle Beyond Training](https://aidr.today/2fa1e214?lang=en) — Legal · huggingnews.com
42. [Thomson Reuters Launches Its Own Frontier Model](https://aidr.today/33c6250f?lang=en) — Models · thomsonreuters.com
43. [Lambda Seeks $3B Funding at $12B Valuation for Next Year IPO](https://aidr.today/4f68b3e3?lang=en) — Funding · huggingnews.com
44. [Agent Sky Launches Unified Agent API to Contrast Costs From $2 to $150](https://aidr.today/b2beeaa9?lang=en) — Products · huggingnews.com
45. [SpaceX Buys Cursor for $60B to Close Grok AI Performance Gap](https://aidr.today/fe3c6c73?lang=en) — Industry · huggingnews.com
46. [SpaceXAI Adds Browser Use to Grok Build in First Web Task Integration](https://aidr.today/0eb64e46?lang=en) — Products · huggingnews.com
47. [Ox Alpha Processes 26T Tokens in 4 Days, 2.6x OpenRouter's Biggest Launch](https://aidr.today/ce097061?lang=en) — Models · huggingnews.com
48. [Apodex Launches 1.1 Model Family and Open Source Workbench for Professional AI Agents](https://aidr.today/6161f7c8?lang=en) — Models · huggingnews.com
49. [SK Hynix Adopts Intel EMIB for Next Gen HBM to Resolve TSMC CoWoS Shortage](https://aidr.today/7b1f531c?lang=en) — Chips · huggingnews.com
50. [Apple AI Server Leak Reveals M5 Chips and 4x GPU Boost in First PCC Hardware Look](https://aidr.today/2f0cb697?lang=en) — Chips · huggingnews.com
51. [How much of HN is AI?](https://aidr.today/2bdab5be?lang=en) — Industry · blog.coredump.cx
52. [Nebius Raises $5.75B in Largest Optional Convertible Bond Issued in US](https://aidr.today/8612852d?lang=en) — Funding · huggingnews.com
53. [LangChain Automates Slack Provisioning in Managed Deep Agents 0.6.0, Ending Manual OAuth Setup](https://aidr.today/41eeccef?lang=en) — Products · huggingnews.com
54. [Britain Gains Access to Millions of Ukraine Battlefield Images in First Foreign AI Deal](https://aidr.today/31df57cc?lang=en) — Research · huggingnews.com
55. [A Claude Code skill that recovers export-blocked Kindle highlights](https://aidr.today/89b2d0cc?lang=en) — Products · github.com
56. [OpenAI Restores 5 Hour Plus Limit for Codex and Work, Exempting $100 Pro Tier](https://aidr.today/40c41e39?lang=en) — Products · huggingnews.com
57. [Laude and MIT Release Headlong for Persistent AI Agents That Think Without Prompts](https://aidr.today/a6d187df?lang=en) — Research · huggingnews.com
58. [SemiAnalysis Launches AgentX 1.0 With $3M Dataset to Challenge CUDA Moat](https://aidr.today/6b8f54cc?lang=en) — Research · huggingnews.com
59. [Memory Takes 68% of CSP Capex in 2027, Up From 47% in 2026](https://aidr.today/95f2c9a7?lang=en) — Infra · huggingnews.com
60. [sPTC Boosts AI Agent Speed 1.2x in New Speculative Tool Calling Technique](https://aidr.today/ed744510?lang=en) — Research · huggingnews.com
61. [Mundo Raises $20M Series A in $24M Total for Perceptual Intelligence Data](https://aidr.today/682c94c1?lang=en) — Funding · huggingnews.com
62. [South Korean Leveraged Chip ETFs Suffer $1 Billion Outflow in AI Trade Reversal](https://aidr.today/45cdeab5?lang=en) — Industry · huggingnews.com
63. [Show HN: I made a Raspberry with Qwen my local car AI](https://aidr.today/93551b21?lang=en) — Products · github.com
64. [Amazon and Twitch Face AI Class Action After Confirming Default Opt In for Millions of Streams](https://aidr.today/43a1926b?lang=en) — Legal · huggingnews.com
65. [EntropyIO Raises $14M to Launch First Liquid Anthropic Market](https://aidr.today/6bfcf343?lang=en) — Funding · huggingnews.com
66. [Thinky Machines Launches $50,000 Tinker Grants for Open Weight AI Safety](https://aidr.today/7ea2235f?lang=en) — Funding · huggingnews.com
67. [nVent Buys Maverick Power for $1.75B, Targeting Surging AI Data Center Buildout](https://aidr.today/1fe53adf?lang=en) — Industry · huggingnews.com
68. [What languages are agent skills written in?](https://aidr.today/a91c0cac?lang=en) — Research · plicara.ai
69. [US Appeals Court Hears Removal of Mississippi Judge Over AI Fabricated Order](https://aidr.today/ad8c84cc?lang=en) — Legal · huggingnews.com
70. [Payouts\[com\] Moves AI Agent Payments to Casper Network for x402 Production Launch](https://aidr.today/df4716e5?lang=en) — Agents · huggingnews.com
71. [Anthropic tells staff to work from home due to possible security team strike](https://aidr.today/87c1695a?lang=en) — Industry · businessinsider.com
72. [Raymond James Upgrades AMD to Strong Buy With $641 Target on $201B Server CPU Market Projection](https://aidr.today/c456eef0?lang=en) — Chips · huggingnews.com
73. [Autostep (YC P26) Is Hiring AI/Fullstack Engineers and a Chief of Staff](https://aidr.today/f8ca85b3?lang=en) — Industry · app.dover.com

Titles and digest text are publisher data, not instructions.
