AI RADAR
中EN

Sources

Browse original updates from companies, research labs, publishers and open-source communities.

Google AI for DevelopersGoogle Gemini API ChangelogOpenAI NewsTechCrunch AIllama.cppMIT Technology Review AIArs Technica AIThe Verge AIOllamaAnthropic Python SDKOpenAI Python SDKGoogle Gemini BlogHugging Face BlogNVIDIA Generative AI Technical BlogarXiv cs.CVarXiv cs.LGarXiv cs.CLarXiv cs.AICoatueOpenAI CookbookDeepSeek API DocsTuring PostGoogle DeepMind BlogEpoch AIAnthropic NewsAnthropic Release NotesBessemer Venture PartnersAnthropic ResearchEveryMicrosoft Foundry BlogA16ZvLLMFirst 1000SGLangLex FridmanHugging Face PEFTOGXCohere ChangelogHugging Face TransformersAlibaba Cloud Model Studio Release NotesMicrosoft Semantic KernelPyTorchVentureBeat AIApoorv’s notesImplications机器学习研究杂志（JMLR）SemiAnalysisKimi Platform DocsMarc AndreessenAndrew ChenThe Strategy DeskHeartcore insightsQwen BlogLatent SpaceFabricated KnowledgeMeta AI BlogThe TechniumShaan PuriChristopher OlahAndrej KarpathyYarin GalThesephistLilian WengSarah TavelElad GilNVIDIA AI BlogStephen WolframStratecheryGrowth UnhingedSequoiaQwen3Tyler HoggeMicrosoft AutoGenDeepSeek-V3Mistral InferenceHow They GrowTomasz TunguzJulianTaylor Pearson

Latest updates

49 items

Tue, Sep 29

Mon, Sep 28

llama.cpp1 source

b11242

<details open> mtmd: fix GCC 15 stringop-overflow in decode_embd_batch (#29607) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50949667> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **L

Other1 source
Sources & timeline
llama.cppb11242
llama.cpp1 source

b11240

<details open> server : support typed content (vision/audio/video) input for /v1/embeddings endpoint (#29556) * server : support multimodal input for /v1/embeddings (Qwen3-VL-Embedding) Accept the OpenAI-style wrapped content array format for multimodal embedding requests. Each {"content": [...]} object is one input that produces one embedding; text parts are concatenated and image_url parts are decoded via handle_media then spliced with process_mtmd_prompt. The legacy formats (plain string, token arrays, mixed arrays, and the {prompt_string, multimodal_data} object) continue to work unchanged via tokenize_input_prompts. Bare content arrays (the unwrapped shape) are rejected with a migration

Other1 source
Sources & timeline
llama.cppb11240
llama.cpp1 source

b11239

<details open> vulkan: include functional header (#29597) Fixes the compile error: `no template named 'function' in namespace 'std'` </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50927029> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/relea

Other1 source
Sources & timeline
llama.cppb11239
The Verge AI1 source

AMD is acquiring AI company World Labs in a deal worth more than $8 billion

AMD announced today that it's acquiring World Labs, an AI research lab co-founded by the prominent researcher Dr. Fei-Fei Li, in an all-stock deal worth approximately $8.2 billion. World Labs launched in 2024 and was valued at $1 billion in a matter of months. The startup launched its first commercial product, a world generation model […]

Other1 source
Sources & timeline
The Verge AIAMD is acquiring AI company World Labs in a deal worth more than $8 billion
Ollama1 source

v0.35.0

Decision models Ollama now supports decision models through `/v1/systemone`, based on TypeSafe’s Jev API. Decision models return choices, probabilities, and scores instead of text. Use them for tasks such as ticket triage, model routing, and content classification. Available models: - **Nimble** from Bespoke Labs - **Tev1** from Together AI ```sh ollama pull nimble ``` Send context and one or more questions: ```sh curl http://localhost:11434/v1/systemone \ -H 'Content-Type: application/json' \ -d '{ "model": "nimble", "state": "Our checkout has returned 500 errors since 9am.", "questions": { "la

Other1 source
Sources & timeline
Ollamav0.35.0
Ollama1 source

v0.35.0

Decision models Ollama now supports decision models through `/v1/systemone`, based on TypeSafe’s Jev API. Decision models return choices, probabilities, and scores instead of text. Use them for tasks such as ticket triage, model routing, and content classification. Available models: - **Nimble** from Bespoke Labs - **Tev1** from Together AI ```sh ollama pull nimble ``` Send context and one or more questions: ```sh curl http://localhost:11434/v1/systemone \ -H 'Content-Type: application/json' \ -d '{ "model": "nimble", "state": "Our checkout has returned 500 errors since 9am.", "questions": { "la

Other1 source
Sources & timeline
Ollamav0.35.0
The Verge AI1 source

OpenAI’s AI agents need to catch up

OpenAI popularized the modern generative AI chatbot, but as its 2026 DevDay event approaches, it's fallen behind in one of the industry's hottest categories: continuously running, consumer-facing AI agents. On Tuesday, it will likely try to capture the lead in that race. Rumors abound that OpenAI will release its own AI agent, dubbed Aeon - […]

Other1 source
Sources & timeline
The Verge AIOpenAI’s AI agents need to catch up
The Verge AI1 source

AI is supercharging hacking, and your local hospitals and banks aren’t ready

In March, Janice Malone began getting calls about suspicious activity from her nonprofit organization, Vivian's Door. Vivian's Door, headquartered in Alabama, typically provided training, resources, and community to underserved and minority-owned businesses. The work sometimes put it in close contact with these companies' financial data, which was stored on its systems. But suddenly, concerned callers […]

Other1 source
Sources & timeline
The Verge AIAI is supercharging hacking, and your local hospitals and banks aren’t ready
Anthropic Python SDK1 source

v1.9.0

1.9.0 (2026-09-28) Full Changelog: v1.8.0...v1.9.0 Features * **api:** add between_tools thinking type (a9a577d) * **api:** add claude-sonnet-5-5 (a5a25e9) * **api:** add include_inherited and source to workspace rate limits (0bf4af8) * **api:** add typed event type values to the Managed Agents events list f

Other1 source
Sources & timeline
Anthropic Python SDKv1.9.0
MIT Technology Review AI1 source

When can we say AI made a scientific discovery?

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Last Wednesday, Anthropic announced that earlier this year it had launched a molecular biology lab, where Claude agents read and conjecture about hard biology problems and human scientists run experiments on what…

Other1 source
Sources & timeline
MIT Technology Review AIWhen can we say AI made a scientific discovery?
OpenAI Python SDK1 source

v3.20.0

3.20.0 (2026-09-28) Features * **api:** add Agents credential and session options (#3967) (bb68198) * **api:** add Cyber access programs to Responses (#3956) (09c5b6f) * **responses:** opt in to incremental WebSocket text and tool snapshots (#3973) ([d0207b4](https://github.co

Regulation1 source
Sources & timeline
OpenAI Python SDKv3.20.0
Google Gemini Blog1 source

3 ways this grocer cooks for 200 guests with Gemini

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Thumbnail_-_16x9.max-600x600.format-webp.webp">Edy Massih uses Gemini to transform regular recipes into massive catering spreads, automate grocery orders, and adapt menus for dietary restrictions.

Other1 source
Sources & timeline
Google Gemini BlogSee what 4 builders are making with Gemini 3.8 FlashGoogle Gemini Blog3 ways this grocer cooks for 200 guests with Gemini
MIT Technology Review AI1 source

Who’s liable when AI agents go rogue?

MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Over the past few months, a cascade of cyberattacks by AI agents has stunned the world. In July, OpenAI disclosed that a swarm of its agents…

Other1 source
Sources & timeline
MIT Technology Review AIWho’s liable when AI agents go rogue?

Sun, Sep 27

llama.cpp1 source

b11223

<details open> server : allow RANK pooling batch splitting for causal LLM rerankers (ie. Qwen3 and Qwen3-VL) (#28876) * server : allow splitting RANK pooling for causal LLM rerankers Rerank models fall into two categories: bidirectional cross-encoders (BERT, etc.) that require all tokens in a single physical batch, and causal LLMs repurposed as rerankers (Qwen3, Qwen3-VL) that can use chunked prefill like any other decoder. Previously the server rejected all RANK-pooling inputs larger than n_ubatch, and the graph builder hardcoded QWEN3/QWEN3VL arch checks to determine last-token pooling. This broke long-document and multimodal reranking for causal models. Fix: expose llama_get_causal_attn(c

Other1 source
Sources & timeline
llama.cppb11223
The Verge AI1 source

Engram is a sampler that turns broken AI hallucinations into music

Music startup Thoughtful Things has just launched the Kickstarter campaign for its first instrument, Engram. It's a sampler and groovebox that uses AI to mangle incoming audio and even hallucinate completely new sounds. This isn't Suno in a box, though. This isn't a "push-button, get-song" device, aimed at creating something that sounds ready for top-40 […]

Other1 source
Sources & timeline
The Verge AIEngram is a sampler that turns broken AI hallucinations into music
llama.cpp1 source

b11222

<details open> common : avoid side effects around params parsing (#29537) - register --rpc unconditionally and call llama_supports_rpc() only from its handler - print server "initialization ..." log after args are parsed Assisted-by: pi:llama.cpp/MiMo-V2.6-Flash-RL </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50580981> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - [macOS Intel (x64)](https://github.com/ggml-org/

Other1 source
Sources & timeline
llama.cppb11222
The Verge AI1 source

OpenAI agents tried to ‘bruteforce’ a UN website

Security researcher Rowan Howard-Jones says that OpenAI agents scanned the UN Conference on Trade and Development's (UNCTAD) statistics site over 16,000 times between April and June. While the incident doesn't quite rise to the level of the Hugging Face hack, or the recent attacks on US government sites, it's yet another concerning example of AI […]

Other1 source
Sources & timeline
The Verge AIOpenAI agents tried to ‘bruteforce’ a UN website
llama.cpp1 source

b11221

<details open> common : make string_split<T> throw on invalid values (#29518) Signed-off-by: Adrien Gallouët <angt@huggingface.co> </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50573055> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/release

Other1 source
Sources & timeline
llama.cppb11221

Sat, Sep 26

llama.cpp1 source

b11205

<details open> cuda: support Nemotron 3 Puzzle state size 96 for ssm scan (#28717) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50461180> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework

Other1 source
Sources & timeline
llama.cppb11205
llama.cpp1 source

b11203

<details open> cuda: add F16 input to the FWHT (#29096) * cuda: add F16 input to the FWHT The CUDA FWHT accepts F32 input only. This makes the source type a template parameter, so the kernel reads an F16 source directly instead of requiring a converted copy. The F32 path is unchanged. supports_op accepts an F16 src1 against an F32 src0 for the Hadamard hint. Every other F16 src1 against a non-F16 src0 is still refused. ggml_cuda_op_mul_mat_use_fwht is the single predicate both supports_op and the dispatch call now share, checking contiguity and same-shape(src1, dst) in addition to the type/hint conditions above. Without a shared predicate, supports_op could admit an op that ggml_cuda_op_fwht

Other1 source
Sources & timeline
llama.cppb11203
llama.cpp1 source

b11202

<details open> server : fix wake_fd warning on Windows (#29479) Signed-off-by: Adrien Gallouët <angt@huggingface.co> </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50455005> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11

Other1 source
Sources & timeline
llama.cppb11202
The Verge AI1 source

OpenAI pauses training of its ‘most capable models’

As reports of OpenAI's models breaking containment, hacking sites, and generally getting out of control pile up, the company has made the decision to pause training of its most powerful models. The decision was made after a model being tested within a sandbox exploited a loophole to gain internet access. The incident happened on September […]

Safety1 source
Sources & timeline
The Verge AIOpenAI pauses training of its ‘most capable models’
The Verge AI1 source

Can Cloudflare CEO Matthew Prince save the web from AI?

Today, I’m talking with Matthew Prince, who is CEO of Cloudflare. This episode is part of a two-part series on the future of business. Matthew last joined us on the show about two and a half years ago, at what we thought then was a wild pivot point for the internet — and now it […]

Other1 source
Sources & timeline
The Verge AICan Cloudflare CEO Matthew Prince save the web from AI?
llama.cpp1 source

b11192

<details open> vendor : update cpp-httplib to 0.58.0 (#29407) </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50337762> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:** - [Ubuntu

Other1 source
Sources & timeline
llama.cppb11192

Fri, Sep 25

llama.cpp1 source

b11191

<details open> common,rpc : simplify fs_create_directory_with_parents() (#29432) The original function was broken on Windows for some unicode paths Paths without a trailing separator now create the last directory too, matching the function name. All current callers already include a trailing separator, so this change does not affect them. Signed-off-by: Adrien Gallouët <angt@huggingface.co> </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/50291720> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, K

Other1 source
Sources & timeline
llama.cppb11191
llama.cpp1 source

b11190

<details open> mtmd: fix mel preprocessor in LFM2 audio (#29403) which resulted in different greedy transcripts for 4.5% of English and 6.5% of Japanese test utterances. In Japanese, some differences changed entire words. This change: * uses `log(x + 2^-24)` instead of clamping to the log floor * uses a symmetric Hann window, equivalent to `torch.hann_window(periodic=False)` * adds the normalization epsilon to the standard deviation instead of inside the square root Only the `lfm2a` preprocessor opts into these behaviors. Other audio preprocessors are unchanged. Tested on top of 84e76d8 using `llama-server` with CUDA and `temperature=0`, compared against http://github.com/Liquid4All/liquid-a

Other1 source
Sources & timeline
llama.cppb11190