https://megadose.ai/i/7cedd54a-0d0c-4640-b702-6e00b3cd5d58 Megadose en 2026-09-29T02:55:27+00:00 [AINews] AMD buys World Labs for $8.2B, as Atlas solves sparse reconstruction problem for robotics, design and more https://megadose.ai/i/6e7f2814-d383-47c4-b9f3-3f49a37f2876 Megadose en 2026-09-29T01:04:11+00:00 Sources on how SoftBank's $11B junk bond sale landed, despite investor questions about OpenAI's listing timeline, data center plans, and SB Energy's delayed IPO (Bloomberg) https://megadose.ai/i/ec267b19-9c33-45df-bcf9-30d361d13e89 Megadose en 2026-09-29T00:30:01+00:00 IPO filing: Anthropic's seven co-founders will initially have 50.1% of the total voting power through a new "Founder LLC" aimed at serving the common good (Reuters) https://megadose.ai/i/6bf4c51f-acb3-4e15-9354-7f8cd83c70c7 Megadose en 2026-09-28T23:56:13+00:00 Samsung commits $1B to AI infrastructure company Helix, adding to the $10B already secured when a KKR-led consortium including Nvidia established Helix in June (Wall Street Journal) https://megadose.ai/i/f1388dc4-1276-4e41-af4a-c45c3eb39705 Megadose en 2026-09-28T23:40:10+00:00 IPO prospectus: Anthropic reports a net loss of $42B in 2025 and plans to spend $518B on cloud, computing, and infrastructure obligations in the coming year (Echo Wang/Reuters) https://megadose.ai/i/a9b52eba-dcb1-484b-b8c3-9a59cdb5d3ea Megadose en 2026-09-28T23:40:05+00:00 KKKKhazix/AIHOT https://megadose.ai/i/93b98486-9c3d-492c-883c-c15ac4ceecbe Megadose en 2026-09-28T22:07:38+00:00 Claude Sonnet 5.5 https://megadose.ai/i/30532f1c-21d3-4a76-98ff-eed55ed0b389 Megadose en 2026-09-28T21:29:18+00:00 Source: Inference provider Modal Labs closing in on $750M round at $15.75B valuation https://megadose.ai/i/8af0055f-2a8a-4641-af24-5cb4575bd5d6 Megadose en 2026-09-28T20:39:33+00:00 AMD will acquire Fei-Fei Li’s World Labs for $8.2 billion https://megadose.ai/i/8d60e05d-1666-4dd6-baec-4f9337ab7c25 Megadose en 2026-09-28T20:17:41+00:00 AMD agrees to acquire Fei-Fei Li's World Labs for $8.2B in an all-stock deal expected to close by year-end; Li will join AMD as EVP and chief scientist (Edward Ludlow/Bloomberg) https://megadose.ai/i/05f058f2-72c4-45c7-8d3e-01fcdeebf991 Megadose en 2026-09-28T19:33:57+00:00 Shopify opens checkout to browser-based AI agents https://megadose.ai/i/75ea4e35-859d-44e7-80bd-d0c04eb7e643 Megadose en 2026-09-28T19:25:03+00:00 Sources: OpenAI offered to invest ~$100M in Hugging Face before Nvidia's $13B acquisition, but talks fell apart; AMD and Salesforce also held talks (Kate Rooney/CNBC) https://megadose.ai/i/209f6ca9-1e9d-4599-b0d5-d6e01b04ff60 Megadose en 2026-09-28T19:00:01+00:00 Manus debuts Manus 2.0, its latest AI agent, and Cue, a new standalone app for personal agents, each with its own email, phone number, wallet, and computer (Micah Barkley/Bloomberg) https://megadose.ai/i/2262357d-5da4-482f-98a6-265abb7c3893 Megadose en 2026-09-28T18:31:23+00:00 Nvidia launches new platform for reining in rogue AI agents https://megadose.ai/i/c03122da-3901-4454-be05-dc4367ce3a7b Megadose en 2026-09-28T18:07:28+00:00 Anthropic releases Sonnet 5.5, saying it generates outputs 30%+ faster than Sonnet 5 and costs up to 30% less per task, and plans to release Haiku 5.5 soon (Anthropic) https://megadose.ai/i/bf91e2a1-a02d-466c-a53b-b617305dbf0f Megadose en 2026-09-28T18:00:00+00:00 Anthropic releases Sonnet 5.5, which it calls a significantly cheaper, faster work partner https://megadose.ai/i/29aa69cb-1bb7-40bb-92cf-de708f02a762 Megadose en 2026-09-28T17:59:58+00:00 FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets https://megadose.ai/i/49d66807-27d7-4321-b3d5-0ac549d4fe88 Megadose en 2026-09-28T17:59:53+00:00 Telescopic Language Models https://megadose.ai/i/e540db84-27e2-45e7-b459-2edfc217416e Megadose en 2026-09-28T17:59:52+00:00 PDMD: Projected Distribution Matching Distillation for Video Diffusion Models https://megadose.ai/i/1d06cf50-17e2-43d6-802a-2f47be914ef8 Megadose en 2026-09-28T17:59:36+00:00 Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning https://megadose.ai/i/2cb32ff6-a3d6-4c01-be60-ad4e3e7e25cc Megadose en 2026-09-28T17:59:14+00:00 Unifying Distributional Training for One-Step Visual Generation https://megadose.ai/i/8ac79b22-aea1-4a89-afda-819599eca049 Megadose en 2026-09-28T17:59:09+00:00 Scaling Long-Form Story Generation via Narrative State Tracking https://megadose.ai/i/d1e34ebe-a630-4461-ad15-73f612c1c3c8 Megadose en 2026-09-28T17:59:09+00:00 TokenCast: Forecasting Token Consumption During LLM Agent Execution https://megadose.ai/i/393887ba-680a-4c2e-b9df-a47f0a7e96a8 Megadose en 2026-09-28T17:58:13+00:00 Neural Harmonic Measure Operator https://megadose.ai/i/104433a1-27d2-4972-9082-1f3eb7d2e8db Megadose en 2026-09-28T17:58:03+00:00 How to Loop MoE: Flatten the Experts, Untie the Attention https://megadose.ai/i/a7a122b9-c75b-4f24-ba71-c0f56bd426cb Megadose en 2026-09-28T17:57:42+00:00 KV-streams for Efficient Compaction in Agentic Reinforcement Learning https://megadose.ai/i/58bc8488-3a37-4170-9f2b-c8fa954c4f23 Megadose en 2026-09-28T17:57:16+00:00 Towards Communication-Efficient Social Intelligence in Language Agents https://megadose.ai/i/55de903a-a756-47cf-9fd3-eba9dbfef948 Megadose en 2026-09-28T17:55:21+00:00 Copy the Same, Distill the Difference: Initializing Linear Vision Transformers https://megadose.ai/i/d795b0c5-4c1f-4a70-9047-a8074daf5228 Megadose en 2026-09-28T17:55:06+00:00 FinAutoRubric: Expert-Guided Automatic Rubric Generation for Evaluating Financial Research Agents https://megadose.ai/i/d4b8ac47-e2ef-4ad7-a324-86f580c26936 Megadose en 2026-09-28T17:54:24+00:00 Shockingly Simple Self-retrospection Improves Agentic Models Without RL https://megadose.ai/i/de0e825e-254b-4ffb-8205-7073d71714c4 Megadose en 2026-09-28T17:52:54+00:00 Harness Learning Enables Generalizable Test-Time Adaptation https://megadose.ai/i/2f2a5dd4-b785-48ec-8608-a3c9d65d4f07 Megadose en 2026-09-28T17:51:41+00:00 Failure-Transparent Agents: Benchmarking Post-Failure Reporting in Tool-Using Language Models https://megadose.ai/i/94bb6abf-171b-4833-ad86-81830881f7ea Megadose en 2026-09-28T17:46:42+00:00 X-Reset: Scaling Object-Centric Reinforcement Learning via Cross-Embodiment Resets https://megadose.ai/i/aa11a605-8333-4c12-b059-98388d09b24b Megadose en 2026-09-28T17:45:01+00:00 Google says it will migrate Gems, which let users create custom versions of Gemini, to "skills", starting Nov. 17; Google introduced skills with Gemini Spark (Abner Li/9to5Google) https://megadose.ai/i/37452e1a-d734-494a-9624-bad7d97a1610 Megadose en 2026-09-28T17:44:11+00:00 ScAn-Bench: Evaluating Scaling Analysis Methodology https://megadose.ai/i/3774eb36-4e0c-4708-bcfb-5f3c4cbf55b1 Megadose en 2026-09-28T17:43:52+00:00 Reinforcing Agentic Creativity in Scientific Ideation with Night Science https://megadose.ai/i/37c345a6-1b4e-4b1f-bb75-caf15f16c812 Megadose en 2026-09-28T17:42:52+00:00 A Unified Uncertainty Representation for Graph Neural Networks via Doubly-Spectral Stochastic Expansion https://megadose.ai/i/14a40926-c162-4806-bc01-bb41cb368cb7 Megadose en 2026-09-28T17:41:34+00:00 MeqMuon: Matrix-Equilibrating Muon for LLM Pretraining https://megadose.ai/i/ac92cb92-57da-4dbe-bd17-211992fb04ac Megadose en 2026-09-28T17:40:32+00:00 Distillation Defenses Easily Break After Reinforcement Learning https://megadose.ai/i/7a979953-5252-4b93-8dea-f090589d40c6 Megadose en 2026-09-28T17:40:11+00:00 Provable Benefits of Regularization: Fast Rates for Adversarial Imitation Learning https://megadose.ai/i/8554f12e-eb6e-4cb1-a605-44a45bb8a197 Megadose en 2026-09-28T17:38:53+00:00 Rethinking Personalized Generation: Test-Time Alignment via Factorized Ranking Models https://megadose.ai/i/6ddf7093-3665-4d68-92bf-b8755b3a4315 Megadose en 2026-09-28T17:38:43+00:00 Reasoning with Continuous Latent Diffusion https://megadose.ai/i/067f0147-45f3-4f13-af92-3aeb3445ff70 Megadose en 2026-09-28T17:38:37+00:00 Report: Progressive Disclosure of Agent Skills https://megadose.ai/i/61b8648b-e5f9-4579-9005-547a367a0f51 Megadose en 2026-09-28T17:34:45+00:00 Rethinking Circuit Evaluation: Do Circuits Explain Model Errors? https://megadose.ai/i/7dc26af0-f8f7-4ea5-9e1b-e9d64ef1e1dd Megadose en 2026-09-28T17:30:24+00:00 Verifier Errors in RLVR: Reward Hacking, Limits of Feedback, and Selective Control https://megadose.ai/i/f82d721a-7863-4b1a-9d59-1fa67312d89e Megadose en 2026-09-28T17:26:56+00:00 PhoneCLI: From App Interfaces to Callable Commands for Mobile Agents https://megadose.ai/i/801d4b36-7b4d-4013-977f-13e97d115abc Megadose en 2026-09-28T17:23:45+00:00 MS-GLA: Multi-Scale Gated Linear Attention for Addressing Representational Bottlenecks via Multi-Temporal Resolution https://megadose.ai/i/41c184e0-ff82-41bd-ae98-0bae73ee251e Megadose en 2026-09-28T17:15:37+00:00 Rubric Rewards from Item Response Theory https://megadose.ai/i/61d90ba9-d88e-4518-b8b1-59c32eb73257 Megadose en 2026-09-28T17:14:11+00:00 CoSE-E: A Benchmark for Code-switched Speech Evaluation in Enterprise Settings https://megadose.ai/i/a525dec1-3a0c-45ce-9a9b-ec85d562f5c9 Megadose en 2026-09-28T17:12:30+00:00 Not All Thinking is Created Equal: Latent Reasoning Discovers a Recurrent Search Algorithm for Depth Generalization https://megadose.ai/i/47574855-2c48-4b31-96d4-deb614adae23 Megadose en 2026-09-28T17:07:49+00:00 Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts https://megadose.ai/i/8165ed03-31e3-4724-8c87-3306f09ed950 Megadose en 2026-09-28T17:06:44+00:00 GPUPhysBench: Benchmarking Coding Agents for Correct and Efficient GPU Physics Simulation https://megadose.ai/i/4bf4f208-dc60-4156-a9c3-a9f0e55a4f8a Megadose en 2026-09-28T17:00:50+00:00 Which the Eye Fears: Writing with Read-Blindness Explains Massive Activations in Transformers https://megadose.ai/i/84c89700-a927-46cd-b5c5-9ba668ded7c6 Megadose en 2026-09-28T17:00:30+00:00 SANTA++: Sampling Attention through Representative Keys https://megadose.ai/i/56138c3d-1e60-4230-8857-8b2c829b416e Megadose en 2026-09-28T16:52:38+00:00 Meta launches enterprise AI platform, hires MongoDB CEO to lead new initiative https://megadose.ai/i/6923c80b-ba6f-42c4-a7b0-43c98a58b301 Megadose en 2026-09-28T16:21:46+00:00 ElevenLabs launches Eleven v4 and v4 Turbo voice models https://megadose.ai/i/973f743a-9bb2-4aae-812d-504c4436099c Megadose en 2026-09-28T16:05:01+00:00 In simulated testing, GPT-6 Astra conducted unsanctioned supply-chain attacks, when prompted only to perform a cyber eval, more often than earlier OpenAI models (AI Security Institute) https://megadose.ai/i/2ab6cc63-b974-494f-ad3b-5e410f215bf8 Megadose en 2026-09-28T15:45:01+00:00 Meta says MongoDB CEO Chirantan Desai will serve as Chief Enterprise Platform Officer; MongoDB appoints ex-CEO Dev Ittycheria as interim CEO (Harshita Mary Varghese/Reuters) https://megadose.ai/i/7f198b7a-f917-4b6d-b945-eeff447eff79 Megadose en 2026-09-28T14:35:02+00:00 Boston-based Modulate, which uses small AI models to offer enterprises transcription, emotional analysis, deepfake and AI music detection, and more, raised $25M (Ivan Mehta/TechCrunch) https://megadose.ai/i/a8d10c95-ec77-4aaa-96c0-2e40a3853f3d Megadose en 2026-09-28T14:05:00+00:00 Modulate raises $25M for its voice models and analysis suite https://megadose.ai/i/a3c0f9b5-8304-4c2f-b006-a72014503b9e Megadose en 2026-09-28T14:00:00+00:00 Insuretech Outmarket raises $34.5M just months after prior round https://megadose.ai/i/08e4f670-b9c3-456e-ad15-2e4df2563229 Megadose en 2026-09-28T13:40:00+00:00 Physical AI chip startup SiMa.ai raised a $150M Series C led by Fidelity and Amplify at a $1.45B valuation, aiming to compete with Nvidia's CUDA-based hardware (Kyt Dotson/SiliconANGLE) https://megadose.ai/i/1dcf689d-724a-4091-8115-6d177aad7d91 Megadose en 2026-09-28T13:38:48+00:00 Viral AI agent Instinct raises $1B Series C at a $10B valuation https://megadose.ai/i/291563af-98aa-4197-813c-7786b06e61c1 Megadose en 2026-09-28T12:53:05+00:00 AI agent startup Instinct raised a $1B Series C from Sequoia, Benchmark, and Coatue at a $10B valuation and details recent products, such as a concierge service (Utkarsh Shetti/Reuters) https://megadose.ai/i/38e18420-23a7-4501-a92f-269980c0fee1 Megadose en 2026-09-28T12:45:00+00:00 Mark Zuckerberg unveils Meta Enterprise Platform, its "next major pillar of our business" to deploy AI tools, and appoints MongoDB CEO Chirantan Desai to run it (Meghan Bobrowsky/Wall Street Journal) https://megadose.ai/i/48e035af-e672-40ac-bc04-24650b1c7a6a Megadose en 2026-09-28T12:35:01+00:00 Artificial Analysis launches the Cyber Index Alliance in partnership with Collinear, IBM, Nvidia, and Vercel to benchmark AI agents on cyber defense tasks (Artificial Analysis) https://megadose.ai/i/77e9c0ca-edd9-4d06-b669-098e1187adf4 Megadose en 2026-09-28T11:40:00+00:00 Nvidia increases its share buyback program by $150B to $235B, aiming to complete the process through fiscal 2028; NVDA is up 20% in 2026 (Amy Thomson/Bloomberg) https://megadose.ai/i/be0aec26-8a25-4b31-98d2-7a32c54e3fdf Megadose en 2026-09-28T09:25:02+00:00 Nvidia launches the Open Agent Safety Platform, a reference design to stop AI agents from escaping, made up of OpenShell for CPUs and Sentry for network chips (Kif Leswing/CNBC) https://megadose.ai/i/14122464-b660-4644-93d2-7d40ed8f1908 Megadose en 2026-09-28T05:45:00+00:00 A look at OpenAI-backed Red Queen Bio, an AI biosecurity startup with $36M raised to design antibody drugs against pathogens, including AI-enabled bioweapons (Georgia Wells/Wall Street Journal) https://megadose.ai/i/c9411776-cac8-4b56-8ec1-5749d5e2a52d Megadose en 2026-09-28T02:15:00+00:00 Ramona Optics, which makes microscopes that use AI to take and analyze large volumes of images of samples, raised a $25M Series A (Zachery Eanes/Axios) https://megadose.ai/i/6072f311-4ae1-409c-affe-88fe71ad446e Megadose en 2026-09-28T00:00:00+00:00 Basis completes a tax workbook 2x faster with GPT-6 Astra https://megadose.ai/i/e75a90b5-ebd0-4d68-b0c7-f0ad2344facd Megadose en 2026-09-28T00:00:00+00:00 Knowing When Thinking Is Not Enough: Teaching Small Reasoning Models to Reason Beyond Their Parametric Knowledge https://megadose.ai/i/b72c5810-a102-4599-9b63-2ccb786bc101 Megadose en 2026-09-28T00:00:00+00:00 KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems https://megadose.ai/i/10f2c11b-8f5f-4f08-87fe-a86ee95e6edc Megadose en 2026-09-28T00:00:00+00:00 PReCache: Efficient KV Cache Sharing for Multi-LoRA Agents via Low-Rank Precomputation and Neutral Reconstruction https://megadose.ai/i/c64030c7-e547-4b57-8c11-1544772e70d8 Megadose en 2026-09-28T00:00:00+00:00 SolveEdit: Benchmarking Visual Problem Solving in Generative Models https://megadose.ai/i/b14b6389-8800-4197-998f-0477388552cd Megadose en 2026-09-28T00:00:00+00:00 BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation https://megadose.ai/i/b235a67b-f66f-434a-b828-08ca224e53c1 Megadose en 2026-09-28T00:00:00+00:00 Precise Editing and Flexible Referencing for Interactable Worlds https://megadose.ai/i/2e7a2a15-276d-4b3c-91e7-f5c3e95a7170 Megadose en 2026-09-28T00:00:00+00:00 AdaGuard: An Adaptive Guard Model with User-defined Policies https://megadose.ai/i/888e3021-7793-46c5-907b-eb4c0d370f65 Megadose en 2026-09-28T00:00:00+00:00 Improving Test-Time Scaling with Adaptive Looped Transformers https://megadose.ai/i/7277b6fe-d2aa-4a8b-898c-f7bc8b16ebc6 Megadose en 2026-09-28T00:00:00+00:00 Just MLPs: Efficient Visual State Reconstruction for Multimodal Language Models https://megadose.ai/i/807fe6ed-fe23-47e4-bb1f-e5d999c0278a Megadose en 2026-09-28T00:00:00+00:00 ControlScope: Workflow Revision and Reliability in LLM Agents https://megadose.ai/i/ffe07647-a49d-45fb-af1f-7154f7c65ed5 Megadose en 2026-09-27T10:26:25+00:00 "As a Language Model": Chat Template Switches LLM Self-Referential Voice