01News
GMKtec EVO-X5 mini PC appears on Geekbench: Ryzen AI Max+ Pro 495 processor
Don't panic! You will be automatically returned to the ITHome homepage in 5 seconds, please wait...
Source links can be verified on the original siteRead original ↗2026-08-15
AI Briefings Daily Report.
An archive of 20 AI news briefs collected on 2026-08-15. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
20 curated selectionsView other dates ↗
02News
Fine-tuning tool-calling large language models: a complete guide based on XYZ-Aquila-SFT and Qwen3
In this tutorial, we implement an end-to-end supervised fine-tuning pipeline for theXYZ-Aquila-SFT dataset, Hugging Face Transformers, PyTorch, and PEFT. We stream and inspect the dataset, parse multi-turn tool-use trajectories, extract structured tool calls, analyze corpus characteristics, and preserve embedded reasoning and observation patterns. We then convert tool schemas between message-embedded and structured formats, render Qwen-compatible ChatML with assistant-only loss masking, prepare a custom PyTorch dataset and collator, and fine-tu
Source links can be verified on the original siteRead original ↗03News
Sony PS5 Pro enhanced PSSR technology revealed: letting AI do less work, but image quality is better
Sony engineers rebuilt the super-resolution technology framework, switched to a new network architecture, restored the traditional reconstruction pipeline, and ultimately improved image quality and temporal stability while also reducing GPU resource usage. #PS5 Pro#
Source links can be verified on the original siteRead original ↗04News
AI-Generated Books Are Flooding Amazon and Reducing Per-Book Income for Human Authors
AI-generated books make up 20 percent of Amazon's self-published catalog but bring in only 12 percent of sales. A new study finds that revenue per book is dropping for human-written titles too, in seven of eight genres. The findings could give copyright plaintiffs the market-harm data their cases against AI companies have been missing.
Source links can be verified on the original siteRead original ↗05News
OpenAI CEO Altman: Two years of college is enough, four years may be a bit long
OpenAI CEO Sam Altman thinks that studying only two years in college was just right for him, and it may also apply to others. "I felt I had already learned a lot in two years, and continuing to study further, the additional gains might become less and less."
Source links can be verified on the original siteRead original ↗06News
Nvidia teams up with six major financial institutions to build 500 billion US dollar AI financing platform; Jensen Huang says some projects may get up to 25% residual value support
Nvidia joins six major financial institutions to form an independent financing platform, planning to mobilize over 500 billion US dollars of third-party capital into AI infrastructure; Jensen Huang introduces up to 25% residual value support to reduce investment risk and promote AI compute as an investable asset
Source links can be verified on the original siteRead original ↗07News
Survey finds young Americans broadly distrust AI, as well as advocate"Embrace AI"tech giant
Young Americans' wariness of AI is deepening, and this distrust also extends to corporate executives and industry advocates who actively promote AI development. The vast majority of respondents said they do not trust AI-investing business leaders such as Musk, Altman, and Zuckerberg.
Source links can be verified on the original siteRead original ↗08News
Agents become"Cheat tool"? Some US students use AI to complete entire online courses on their behalf
AI cheating has escalated from ghostwriting assignments to almost "attending classes on behalf of students," exposing the difficulty of preventing cheating in online education compared to physical classrooms. Offline courses can verify student mastery through exams and oral presentations, whereas online courses struggle to employ similar methods.
Source links can be verified on the original siteRead original ↗09News
Up 32% month-on-month in July, OpenAI's enterprise AI business revenue has surpassed consumer business
CNBC published a blog post yesterday (August 14) reporting that OpenAI CFO Sarah Friar revealed to shareholders this Friday that OpenAI's enterprise business revenue has surpassed its consumer business dominated by ChatGPT.
Source links can be verified on the original siteRead original ↗10News
Anthropic Prepares"Largest in history"IPO, estimated 2028 revenue may reach 190 billion to 200 billion US dollars
Compared with Anthropic's annualized revenue run rate of 47 billion US dollars (currently about 317.515 billion RMB by exchange rate) announced this May, the 2028 target scale is several times higher, also meaning investors need to bet big on high-speed growth over the next two years.
Source links can be verified on the original siteRead original ↗11News
Plaintiff hides invisible AI instructions in court filing, attempting to covertly influence automated review
A plaintiff in Connecticut embedded invisible prompt injections in court filings, formatted in 3-point white text on a white background, to manipulate a potential AI review system. Judge Spader compared the attempt to secretly tampering with a jury and revoked the plaintiff's electronic filing privileges. The court stressed that Connecticut doesn't use AI to review filings, but the intent alone was enough to warrant sanctions.
Source links can be verified on the original siteRead original ↗12News
Anthropic shares six cost-saving tips for Claude Code: managing token cache can reduce costs by tenfold
Official estimate: developers use Claude Code with average daily Token consumption of 13 US dollars, and different questioning methods differ in cost by several times. Anthropic officially summarized six money-saving tips, the core being reasonable use of caching and streamlined context, helping developers save large expenses
Source links can be verified on the original siteRead original ↗13News
Anthropic's Q2 2026 revenue exceeds $11.5 billion, achieving operating profit for the first time
Bloomberg today (August 15) published a blog post reporting that Anthropic told potential investors that the company's Q2 2026 revenue exceeded 11.5 billion US dollars, a year-on-year increase of more than 14 times, and it achieved operating profitability for the first time.
Source links can be verified on the original siteRead original ↗14News
After Anthropic Adds Watermarks to Claude-Generated Text, Some Users Worry"Trace"Unsubscribe
After Anthropic announced adding watermarks to text generated by Claude, some users began opting out. Subsequently, Business Insider interviewed four users who had already unsubscribed from Claude. They worried that long-lasting markers would remain in their work products and might expose their use of AI in client projects such as code.
Source links can be verified on the original siteRead original ↗15News
OpenAI sets up a dedicated mailbox; employee complaints can reach CEO Sam Altman directly
Fortune published a blog post yesterday (August 14), reporting that OpenAI has an internal mailbox named friction, where employees can directly feedback work obstacles to management including CEO Sam Altman and President Greg Brockman.
Source links can be verified on the original siteRead original ↗16News
World first: human employee fired for being late 17 times across 23 shifts by"AI Boss"Claude fires
Time magazine published a blog post today (August 15), reporting the world's first case of using AI as a corporate manager, who ultimately decided to fire its employees.
Source links can be verified on the original siteRead original ↗17News
Tencent QQ Bot officially announces integration with DeepSeek Harness: supports one-on-one chat / group chat, integration requires only 3 steps
Tencent QQ posted on Weibo today, announcing that QQ Bot robots have integrated with DeepSeek Harness and official plugins. After integration, QQ Bot robots possess AI agent capabilities, supporting both one-on-one and group chats.
Source links can be verified on the original siteRead original ↗18News
New benchmark confirms AI models still perform poorly at visual perception
Moonshot AI's PerceptionBench tests how well multimodal AI models can actually "see," separate from logical reasoning. No frontier model reaches 60 percent accuracy, and GPT-5.6 Sol leads by a narrow margin. Many supposed reasoning errors actually happen as early as the image-reading stage.
Source links can be verified on the original siteRead original ↗19News
Paper: Rational adoption of AI may or may not destroy the collective professional capability of an entire occupational field
A new research paper frames AI adoption as a "tragedy of the cognitive commons." Every company that cuts entry-level jobs benefits individually, but the collective expertise of entire professions erodes. The consequences may not become visible until 2030 to 2045, when today's missing junior talent should have become tomorrow's experienced workforce.
Source links can be verified on the original siteRead original ↗20News
Stop sending thousand-line giant PRs: a code reviewer's gripe in the AI era
I'm tired boss. I'm tired of reviewing one, two, three thousand line PRs because some agent was able to "one shot the whole issue." Small PRs were never asked for because they're easier to write, it's always been for the benefit of the reviewer. AI is such a boon to the industry, but it's becoming such a liability for reviewers and maintainers. Maybe I'm just an old man yelling at the clouds, but please, stop sending me huge PRs.
Source links can be verified on the original siteRead original ↗