Create

Sign in to ReadmeX

or
ReadmeX
ReadmeX

A clearer picture, in a conversation.

Catch up on what matters, then ask a little deeper.

Your community and people briefings stay personal to you.

AI News

‹Thu, October 8›
All dates
2415
Industry

Xitel New Energy approves listing transfer, capital raise to tap AI data center energy

Taiwanese energy-storage company Xitel New Energy (熙特爾新能源) held an extraordinary shareholders' meeting on Oct. 8, approving a cash capital increase tied to its application to move its listing to the main board, and elected a new director: Gambo Long, Asia-Pacific vice president and Taiwan general manager at Astera Labs. The company said the move targets AI data center energy demand and that it is developing backup battery units (BBU), high-voltage battery systems and smart energy management for data centers shifting toward higher power density and 800V high-voltage DC supply. General manager Chen Po-hsun said Taiwan's AI server manufacturing, system integration and power supply chain support next-generation data center power infrastructure.

科技新报·
2424
Industry

GitHub tops 225M users as non-coders flood in

GitHub grew from roughly 100M users in early 2023 to 180M in September 2025 and past 225M by July 2026, according to Microsoft data cited by the article — growth that cannot be explained by professional developers alone. The piece argues that AI-lowered barriers have drawn in large numbers of non-technical users, spawning popular non-code projects such as classical Chinese poetry datasets, a "high-value life guide" that passed 27,000 stars in under a month, prompt libraries and "colleague.skill." GitHub itself has acknowledged that AI is driving interest in software development but said it "cannot fully explain the surge."

36氪 人工智能·
2434
Industry

Stack Overflow survey: AI coding drives burnout and career anxiety

A Stack Overflow survey of 30,903 developers across 169 countries found nearly half feel disengaged with their work, roughly a third say they are unhappy, and only slightly more than a fifth are satisfied; among the disengaged, burnout and tech fatigue were the leading cause at about 21%. AI tooling is already mainstream: close to two-thirds use AI coding assistants, 62% use general-purpose AI chat tools, and 31% of daily AI users spend more than four hours a day with it. Trust lags adoption: fewer than 10% would trust AI for important decisions, and about 80% say source attribution matters when reviewing AI-generated code or answers; a chart cited in the report puts developer trust in AI output at 87% this year, up sharply, though 48% trust it only for low-risk, easily verifiable tasks.

科技新报·
2445
Industry

Shangmei hotel report: Doubao channel bookings up 232% over National Day holiday

Shangmei Digital Intelligence Hotel Group released its 2026 National Day small-city travel vitality report, saying average daily rate rose 0.8% year on year and revenue per available room 0.5% year on year during the Oct 1–7 holiday. The report said that as AI and other new entry points spread, sales via the Doubao channel grew 232% from the previous period, orders rose 40.5% and the number of stores receiving orders rose 61.2%.

36氪快讯·
2454
Industry

CITIC Securities: AI industry focus shifting to inference and monetization

A CITIC Securities research note says recent proposals by model makers such as Anthropic and OpenAI to slow frontier capability gains triggered sharp market swings and wide debate. The brokerage argues the move is less about founders' moral qualms than about shrewd business calculation and operating pressure at leading model makers, with rising regulatory costs actually favouring incumbents. It expects regulation to improve gradually and top firms to focus more on deployment to ease financial pressure, shifting the industry's centre of gravity toward inference and monetization, and remains bullish on CSPs, platform software vendors and selective hardware/semiconductor opportunities.

36氪快讯·
2465
Industry

Ex-Omidyar Network CEO launches Andaris, an AI investor network

Mike Kubzansky, former CEO of philanthropic investment firm Omidyar Network, is launching Andaris, a group for large investors meant to facilitate information sharing about the AI industry and foster more trustworthy practices. The report did not disclose the network's backers, membership or operating details.

The Information·
2474
Models

Why isn't the industry freaking out about DeepSeek 4.1 Flash?

A developer blogger says a month of heavy use of DeepSeek 4.1 Flash left him unable to tell it apart from frontier models such as Claude Opus 5.5 in everyday coding, at a fraction of the cost; he sometimes uses Opus for a final code review and has DeepSeek apply the fixes. He also credits a roughly 437x smaller KV cache versus DeepSeek V1 for keeping all-day sessions under a dollar, and dismisses the dispute over Chinese labs allegedly distilling Claude's training data as irrelevant to developers chasing value.

Hacker News · AI（100+ 分）·
2484
Industry

Taiwan dethrones Korea as top AI trade on supply-chain breadth

TechNews reports Taiwan's weighted index beat Korea's KOSPI by 23 percentage points in July–September, the widest gap since the early 2000s, with year-to-date returns of about 72% versus 65%. A September Bank of America fund manager survey found nearly 40% of respondents overweight Taiwan versus 25% for Korea, and 35% see Taiwan as the main beneficiary of the next AI cycle against just 5% for Korea. M&G Investments portfolio manager Vikas Pershad said Taiwan's earnings growth comes from volume expansion and is therefore broader-based, while Korea's upgrades are concentrated in price factors and rely heavily on Samsung and SK Hynix.

科技新报·
2494
Industry

Toshiba to spend $380M expanding AI storage HDD output

Toshiba is investing 60 billion yen (about $380 million) to expand hard-disk drive manufacturing in the Philippines — its first major HDD expansion in roughly five years — aiming to nearly double capacity for AI and cloud storage drives by fiscal 2027. New lines will also build drives holding about 40% more data than current products, with 65TB-class drives planned around 2030 and 100TB models later. Rival suppliers are also stretched: Seagate has reportedly allocated most nearline HDD capacity through 2028, and Western Digital has committed much of its 2027 output.

TechRadar·
2505
Industry

C.H. Robinson to buy RXO in $5.8 billion AI-driven freight deal

Bloomberg Technology reports that freight broker C.H. Robinson (CHRW) is acquiring RXO in a $5.8 billion deal it frames as AI-driven. The report says AI motivated the size of the move, but gives no detail on the technology, deal terms or financing.

Bloomberg Technology·
2514
Open source

Google open-sources AQuA, an agent that diagnoses production AI agents

Google's developer blog introduces AQuA (Ambient Quality Agent), released as composable building blocks in the google/adk-recipes repo. It runs alongside your agent in your Google Cloud project, sampling up to 1,000 recent sessions from Cloud Trace, Cloud Logging or BigQuery on a schedule, after deployments or on demand, and pushes them through a five-stage pipeline (sample, review, cluster, verify, track) that yields structured findings; it never sits in the request path and does not edit code or open pull requests itself. In a demo sweep of 32 travel-concierge sessions, it produced 42 findings and 9 candidate clusters, of which the Gemini 3.7 Flash verifier rejected 3 false positives, leaving 6 verified issues.

Google Developers Blog·
2524
Products

Hugging Face's ML Intern builds six small models in days on a tiny budget

In a Hugging Face blog post, author yvrjsharma describes using the ML Intern agent inside HuggingChat to build six models in a couple of days, including distilling Qwen-Image 2.1's 9B prompt rewriter into a 0.8B 'Pocket' version that runs on CPU (99.7% valid output, about a quarter of the teacher's tokens) and fine-tuning Qwen3.5-2B for citrus-disease diagnosis, which rose from 14.9% to 52.8% accuracy on 335 test images. The agent plans the work, requests a budget before spending, runs a smoke test, then trains, evaluates and publishes to the Hub; the author puts the first project's total compute at about USD 16. The post also shares prompting advice, such as asking for a pre-training baseline and capping total spend.

Hugging Face·
2534
Research

Apple's Normalizing Trajectory Models generate images in four steps with exact likelihood

Apple Machine Learning Research published Normalizing Trajectory Models (NTM) at NeurIPS, which models each reverse diffusion step as an expressive conditional normalizing flow with exact likelihood training, pairing shallow invertible blocks per step with a deep parallel predictor across the trajectory and trainable from scratch or from pretrained flow-matching models. Its exact trajectory likelihood also enables self-distillation: a lightweight denoiser trained on the score function induced by the model itself produces high-quality samples in four steps. On text-to-image benchmarks, NTM matches or outperforms strong image generation baselines in just four sampling steps while uniquely retaining exact likelihood over the generative trajectory.

Apple Machine Learning Research·
2545
Industry

Taiwan minister: Taiwan firms' US investment aims to 'define the market' in a decade

Taiwan's Minister of Economic Affairs Kung Ming-hsin said Taiwanese companies are shifting their US investment rationale away from lowering production costs toward getting closer to the US market through supply chains, moving from "understanding the market" to "defining the market." He said Taiwan is currently at the front of AI-related industries with a hard-to-replace AI supply chain, and that the US wants Taiwanese firms to invest there and form supply-chain partnerships. The government hopes three to five years of overseas deployment will bring closer market integration, with results visible in about ten years.

科技新报·
2555
Industry

Three Chinese AI companion apps go overseas: Popia, MochiChat, imely

A Xia Guang She piece published via 36Kr surveys the overseas push by Chinese AI social apps: per Appfigures Intelligence, 337 revenue-generating AI social apps were operating worldwide as of July 2025, 128 of them launched in 2025 (about 38%), with 2.2 billion cumulative downloads and $221 million in consumer spending. It profiles three products from Chinese teams launched in 2026: Popia (launched January, US-focused), which turns character creation into a multi-turn AI conversation and has taken in roughly $400,000 in mobile revenue while entering the top 25 of US entertainment grossing charts; MochiChat (April), focused on Southeast Asia with nearly 1.2 million users and about $120,000 in revenue, adding virtual transfers, a social feed and five intimacy levels; and imely (April), which has passed 500,000 users in Vietnam and lets relationships deteriorate via a "negative intimacy" mechanic. The author argues that amid fading traffic dividends, heavy imitation and rising compliance barriers, newcomers can still win by targeting narrower markets and needs.

36氪 人工智能·
2565
Industry

AI recruiting startup Dex hires Secret Escapes and Progression founders

Dex, a London-based AI recruitment platform for software engineers, has appointed Tom Valentine, co-founder of travel deals site Secret Escapes, as chief operating officer, and Jonny Burch, co-founder of career development platform Progression, as head of design. Valentine will lead commercial work and grow the core marketplace matching engineers with employers, while Burch will build the design team and lead the candidate-facing product. Dex's AI "talent agent" talks to engineers about their skills, experience and goals before matching them with jobs, and runs on a mix of models from Google, Anthropic and OpenAI. The company raised a $5.3m seed round led by Notion Capital in April, according to Tech.eu.

The Next Web·
2575
Industry

Teen's Claude-planned mountain hike ends in helicopter rescue

A 16-year-old, Bryce Vincent Gowryluk, used Anthropic's Claude to plan a new route to the summit of Crown Mountain near Vancouver and ended up stranded on the "Widowmaker Arete," a steep cliff requiring climbing gear, according to the New York Times. He called police from a five-foot-wide ledge, and the volunteer North Shore Rescue team tracked his phone, dropped two members by helicopter and winched him to safety. The teen said he doesn't blame Claude and takes responsibility, noting he also used Google Maps and AllTrails; rescue manager Paul Markey said Claude has no real knowledge of locations or terrain and that AI is no substitute for experience and common sense.

The Decoder·
2585
Industry

HP survey says AI is creating jobs in UK businesses

An HP survey of nearly 20,000 desk-based workers worldwide found 46% of Britons say AI has positively affected their lives, and 52% use employer-provided AI tools daily or weekly, up from 38% a year earlier. Three in five UK business leaders say AI adoption has created new roles or whole teams, while just 10% say it is mainly replacing or reducing roles. Agentic AI use stands at 45% among UK desk-based workers, though it falls with seniority — 38% of knowledge workers versus 56% of business leaders and 71% of IT decision-makers — and 38% of knowledge workers still fear AI could replace their role, with 37% fearing replacement by someone more AI-savvy.

TechRadar·
2595
Industry

Hongguo hits 168M daily users as AI micro-drama memes shape language

A commentary republished by 36Kr says ByteDance's free short-drama app Hongguo has reached 168 million daily active users with about 125 minutes of use per person per day, and cites figures showing roughly 128,000 new short dramas launched across the web in Q1, about 95% of them AI-generated animation or AI-live-action dramas; in July, AI-animated titles took seven of the top ten slots on Hongguo's popularity chart. The author argues that lines like "吃了不烧心" read as overwrought and formulaically parallel because large models have no bodily experience, yet get reinforced through sheer volume until they enter everyday speech. The piece ends by suggesting that instead of only teaching AI to sound human, people may gradually adapt to AI's language habits.

36氪 人工智能·
2605
Open source

vLLM climbs GitHub Trending among Python projects

vLLM is a high-throughput, memory-efficient inference and serving engine for large language models, originally developed in the Sky Computing Lab at UC Berkeley and now maintained by a community of more than 2,000 contributors. It manages attention key-value memory with PagedAttention and supports continuous batching, chunked prefill, prefix caching, speculative decoding and quantization formats including FP8, INT8/INT4, GPTQ/AWQ and GGUF, alongside an OpenAI-compatible API server. The engine runs on NVIDIA, AMD and Intel GPUs plus x86/ARM/PowerPC CPUs, and supports 200+ model architectures on Hugging Face.

GitHub Trending · Python·
Next →← Previous