https://megadose.ai/i/da4bc3f6-f8e5-48d2-8004-26e0bc5aaa8e Megadose en 2026-09-22T18:40:01+00:00 Sources: Mirendil, founded by former Anthropic researchers to build self-improving AI, is in talks to raise ~$1B led by Kleiner Perkins at a $5B valuation (Bloomberg) https://megadose.ai/i/ff82dd8c-28d7-40a0-86f8-1d4eb0c64751 Megadose en 2026-09-22T18:15:08+00:00 OpenAI launches GPT-6 Sol and Luna, saying Sol makes about half as many mistakes as GPT-5.6 Sol and Luna matches GPT-5.6 Sol's performance at ~1% of the cost (David Gewirtz/ZDNET) https://megadose.ai/i/7532509b-8669-45e2-a145-545ef4e5d66c Megadose en 2026-09-22T18:00:00+00:00 OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes https://megadose.ai/i/2412752c-4b20-49c6-b77b-17c6ac4938d2 Megadose en 2026-09-22T18:00:00+00:00 Introducing GPT-6 Sol and Luna https://megadose.ai/i/8be3bf9b-4da6-4c12-99fa-1e4bebedec60 Megadose en 2026-09-22T16:42:22+00:00 Anthropic says Opus 5.5 matches Fable 5.1 "on most tasks" while costing about 40% less to run than Opus 5; Opus 5.5 costs $4/1M input and $20/1M output tokens (Matthias Bastian/The Decoder) https://megadose.ai/i/6f9d545f-c31e-42b1-88e2-14ae0534e291 Megadose en 2026-09-22T16:30:07+00:00 Anthropic releases Opus 5.5 with lower prices and Fable-level performance https://megadose.ai/i/5c1d34fe-f608-4b98-8952-0d386db1f2ca Megadose en 2026-09-22T16:27:30+00:00 Claude Opus 5.5 https://megadose.ai/i/1adc3b7e-cbb2-4eaf-aec3-a989e866e1f4 Megadose en 2026-09-22T15:35:00+00:00 Chamelio, which is building AI tools designed to execute routine legal workflows and escalate only the cases that require human judgment, raised a $26M Series A (CTech) https://megadose.ai/i/04275fb4-ceb2-46ed-979a-b7cff0716587 Megadose en 2026-09-22T14:50:01+00:00 Snorkel AI, which offers an "agentic data development platform" with humans and AI agents to create and vet data, raised $350M at a $3.5B valuation (Krystal Hu/Reuters) https://megadose.ai/i/7e0b78d0-0271-494f-a933-0d8076c936aa Megadose en 2026-09-22T12:50:01+00:00 Alphabet-owned robotics software company Intrinsic open-sources Intrinsic Core under Apache 2.0, giving developers building blocks for physical AI systems (Mike Wheatley/SiliconANGLE) https://megadose.ai/i/415d8639-c0b3-4554-b3ea-91ec6bfa5cb6 Megadose en 2026-09-22T12:45:02+00:00 Australian healthcare AI startup Heidi raised $340M, including a $100M Series C led by Blackbird at a $900M valuation, taking its total funding to $430M+ (Olivia Poh/Bloomberg) https://megadose.ai/i/0d3bc30a-8280-496e-847e-8b43f0418805 Megadose en 2026-09-22T12:40:00+00:00 Baselayer, which uses AI to help financial institutions verify businesses and assess fraud risk, raised a $35M Series A led by M13, for ~$40M in total funding (Mary Ann Azevedo/Crunchbase News) https://megadose.ai/i/1d9d5bfa-f61e-49c7-9019-ecdbea131bf9 Megadose en 2026-09-22T12:25:05+00:00 Cisco Talos releases CAIRN, an open-source framework designed to classify and analyze AI-integrated malware by tracking AI metadata and behavioral fingerprints (Lily Hay Newman/Wired) https://megadose.ai/i/59060bb9-5b51-4833-a5ad-1077cc06aa5f Megadose en 2026-09-22T12:23:05+00:00 Nscale’s IPO will test Wall Street’s appetite for concentrated AI bets once again https://megadose.ai/i/0b64c676-6dc4-448f-84a2-f958aba317a8 Megadose en 2026-09-22T12:15:01+00:00 Cognex agrees to acquire robotics and 3D computer vision company RealSense in a ~$600M deal, including $500M in cash; Intel spun off RealSense in July 2025 (Sophie Shulman/CTech) https://megadose.ai/i/6d070638-66b9-40a2-b5e1-5ebb0728c6c5 Megadose en 2026-09-22T11:15:01+00:00 The Austrian Academy of Science, Mistral, and Sail Reply plan to launch Apollo, an Ancient Greek LLM trained on ~600M historical Greek words, available for free (Joel Khalili/Wired) https://megadose.ai/i/e992102e-25ff-4db1-be0a-de1be78c4b50 Megadose en 2026-09-22T10:55:01+00:00 US startup Firebird is building a 300MW data center in Armenia, set to host 70K+ Nvidia chips by 2027; sources: Trump pledged Nvidia export approvals to Armenia (Mackenzie Hawkins/Bloomberg) https://megadose.ai/i/28037548-3944-4e40-aebe-88df5abefb03 Megadose en 2026-09-22T10:20:02+00:00 Finnish neocloud startup Verda raised $189M led by Emergence at a $1B+ valuation and aims to raise up to $1.5B in equity and debt in 2026 and up to $10B in 2027 (Paula Doenecke/Bloomberg) https://megadose.ai/i/35d669cc-ec80-4ea5-ad01-9b6314d4d485 Megadose en 2026-09-22T09:50:00+00:00 Chinese optical transceiver maker Ligent jumps 19% in its Hong Kong debut, giving it a ~$4.6B market cap, after raising ~$727M, the latest Hong Kong AI IPO (Sangmi Cha/Bloomberg) https://megadose.ai/i/a22d8a94-a55a-4df2-8d48-d804c06c39f1 Megadose en 2026-09-22T09:36:31+00:00 Verda (Finland) raises $189M in Series B https://megadose.ai/i/e52f671b-3fe5-46c6-a3a3-cc86090eae09 Megadose en 2026-09-22T06:30:56+00:00 [AINews] Xiaomi MiMo-V2.6-Pro 1T-A42B: the new top Open Weights model, trained for $3M https://megadose.ai/i/66f754aa-541b-4836-baed-0ca230ddddc2 Megadose en 2026-09-22T05:50:02+00:00 Belgium-based AI recruitment platform Spott raised a $21M Series A led by Balderton, with Base10, YC, and Fortino participating, bringing total funding to $24M (Chris Metinko/Axios) https://megadose.ai/i/e528123c-7c94-4579-9d39-dc6a6216b8cd Megadose en 2026-09-22T04:20:01+00:00 Tencent launches Hy Image 3.5 Preview, saying during internal testing the image model achieved results on par with ByteDance's Seedream 5.0 Pro (Bloomberg) https://megadose.ai/i/2b3bdcbe-203e-4a2c-9a06-adfa31a38106 Megadose en 2026-09-22T03:30:01+00:00 Alibaba CEO Eddie Wu says the company plans to train a 5T- to 10T-parameter AI model, as it lays out a sweeping push across AI models, chips, and data centers (Reuters) https://megadose.ai/i/6ae0ba35-e7e1-4332-b1df-67ee08ee9735 Megadose en 2026-09-22T03:25:00+00:00 Alibaba's T-Head unveils the Zhenwu V900 AI accelerator, which it says triples its predecessor's performance and can scale to clusters of up to 500,000 units (Bloomberg) https://megadose.ai/i/c717df96-9338-4445-83a6-006fdb60c2f4 Megadose en 2026-09-22T02:15:00+00:00 MiMo-V2.6-Pro ties Grok 4.7 (xHigh) and beats GLM-5.3 (max) on Artificial Analysis' Intelligence Index, making it the benchmark's top-scoring open-weight model (Carl Franzen/VentureBeat) https://megadose.ai/i/ac53f7e2-bbe6-4002-845b-c6b79cc0f044 Megadose en 2026-09-21T23:55:00+00:00 Shopify plans to allow Meta's Muse to complete purchases on behalf of users at Shopify-powered stores via its one-tap checkout service Shop Pay (Kelly Cloonan/Wall Street Journal) https://megadose.ai/i/01d104de-353e-48cd-9916-d2e351a7a823 Megadose en 2026-09-21T23:14:40+00:00 Xiaomi open-sources MiMo-V2.6 Pro and Flash models https://megadose.ai/i/11d7f471-380b-47f0-beac-22795c64c927 Megadose en 2026-09-21T23:09:20+00:00 Jev introduces a new shape of LLM - System One, aka Decision Models https://megadose.ai/i/c09ece7a-66e9-40c9-b76f-c92e67d4c83f Megadose en 2026-09-21T22:32:03+00:00 SpaceXAI releases Grok 4.7 for coding and knowledge work https://megadose.ai/i/e089b173-c445-4c0d-8dcf-9076827ca80f Megadose en 2026-09-21T22:10:02+00:00 Xiaomi debuts open-weight omnimodal models MiMo-V2.6 Pro and Flash; Pro allegedly performs "on par with Opus 5 and GPT-5.6 Sol across most agent benchmarks" (Xiaomi) https://megadose.ai/i/d18db9c8-00e0-4f64-9c0d-732b2df6cb35 Megadose en 2026-09-21T19:50:01+00:00 Sources: SoftBank's SB Energy has delayed its IPO, originally planned for this month, as investors question the company's sought-after valuation of $50B+ (Maureen Farrell/New York Times) https://megadose.ai/i/9d6b92e3-780c-40e8-8ee5-e281fe3ccb94 Megadose en 2026-09-21T19:05:00+00:00 Sensor Tower: Muse was downloaded 902K+ times in the six days after its launch on Sept. 8, vs. Meta AI's 773K in the same post-launch period; META jumps 12%+ (Bloomberg) https://megadose.ai/i/b0572d15-3bb7-42df-956e-ac907952ff63 Megadose en 2026-09-21T17:59:33+00:00 GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay https://megadose.ai/i/bf88a66d-fbe0-4069-a74f-434f7273ce1d Megadose en 2026-09-21T17:57:20+00:00 Critical-State RL: Diagnosing Trainable States for Multi-Turn Tool Use https://megadose.ai/i/affc7fc9-1004-4431-b626-1be7ebacf5c4 Megadose en 2026-09-21T17:57:06+00:00 WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory https://megadose.ai/i/72e7a114-08fd-49fb-91e0-9cde90c787fe Megadose en 2026-09-21T17:56:53+00:00 onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction https://megadose.ai/i/4e944507-65ca-4b29-97f0-5bc89ae5ecc2 Megadose en 2026-09-21T17:55:48+00:00 LoRA-generating hypernetworks for efficient on-device LLM generative personalization https://megadose.ai/i/95dcabed-c1d2-4db0-a39f-fa073d76c8c6 Megadose en 2026-09-21T17:55:24+00:00 DexTacWAM: A Visuo-Tactile World-Action Model for Dexterous Manipulation https://megadose.ai/i/c7e36fe3-86f5-442d-b564-f54abdc28b7d Megadose en 2026-09-21T17:54:49+00:00 RRSI: Regularized Recursive Self-Improvement of Agent Harnesses https://megadose.ai/i/7d14f38b-de38-49c0-a7fe-40a170800303 Megadose en 2026-09-21T17:54:35+00:00 DolphinBench: Mapping the Pareto Frontier of Agent Memory https://megadose.ai/i/ac1aa6c7-4cb2-4934-8261-81c560553d18 Megadose en 2026-09-21T17:53:38+00:00 Rare Event Estimation via Iterative Unalignment https://megadose.ai/i/43ff4341-7f0b-40cc-a31e-877fbe2019b1 Megadose en 2026-09-21T17:52:48+00:00 Emergent Collusion in Long-Horizon LLM Agent Interaction https://megadose.ai/i/de9b1c2f-4775-4e7d-8812-7171c67e8913 Megadose en 2026-09-21T17:46:25+00:00 Generative Tutorial: Towards Live Contextualized Visual Instructions for Physical Tasks https://megadose.ai/i/6b35c31c-fcbf-4dd7-8bcc-a2926b9cdf7b Megadose en 2026-09-21T17:42:17+00:00 Learning Physics from an Imperfect Ancestor https://megadose.ai/i/de7f19f5-54fc-46d5-b291-9153717a6ae5 Megadose en 2026-09-21T17:39:51+00:00 Exactness at Inference: A Representational Criterion for Out-of-Distribution Generalization https://megadose.ai/i/7a193bff-0618-4205-bf12-81ca21bfc70a Megadose en 2026-09-21T17:22:44+00:00 Et Tu, Brute? Economic Misalignment in Personal AI Agents https://megadose.ai/i/8e9e748b-91f0-412a-a581-26ce6ad41663 Megadose en 2026-09-21T17:18:19+00:00 BackTrend: Evaluating Scientific Weak-Signal Prediction via Backward Reconstruction https://megadose.ai/i/58e35edb-64df-462a-a7ec-4c3ce7658619 Megadose en 2026-09-21T17:00:23+00:00 Amazon blocks Meta’s new Muse AI agent from shopping on amazon.com https://megadose.ai/i/187057e8-0e11-45b7-bc4b-258efba7bce6 Megadose en 2026-09-21T17:00:15+00:00 Human-LLM Deliberation as Interactive Proof: Conditions for Verifiability Without Transparency https://megadose.ai/i/75cc3800-c98f-4855-834f-a81903d6b695 Megadose en 2026-09-21T16:59:06+00:00 SLICEChat: Progressive In-Encoder Token Pruning for Whole-Slide Pathology Language Models https://megadose.ai/i/c4fc1147-5f49-4c3c-8975-8580432df560 Megadose en 2026-09-21T16:55:24+00:00 OSWorld-Pro: Process-based Evaluation for Computer Use Agents https://megadose.ai/i/d107b908-a431-4fff-848b-066e87b24b06 Megadose en 2026-09-21T16:53:00+00:00 The Copy Ceiling: An Input-Exposure Control for Ontology-Grounded Generation over Curated Corpora https://megadose.ai/i/589918a5-507b-4c0f-b706-7e663ebd12b6 Megadose en 2026-09-21T16:48:50+00:00 Pinocchio: Fast Uncertainty Estimates for Black-Box Language Models https://megadose.ai/i/b7e57738-0f80-46bd-97be-7ca218d7ecd1 Megadose en 2026-09-21T16:35:57+00:00 SE(3) Neural Potential Fields for 6-DoF Trajectory Planning Directly from Images Without Explicit 3D Reconstruction https://megadose.ai/i/089a5968-578a-4e24-96fb-c3fd8e43e6df Megadose en 2026-09-21T16:25:24+00:00 SPECTRA: Adaptive Execution of Speculative Decoding on a Runtime-Reconfigurable Tiled Architecture https://megadose.ai/i/71cae303-872c-4772-9fdb-ed3baa3d03ea Megadose en 2026-09-21T16:20:25+00:00 PredActor: Predictive Action Diffusion for Steerable Onboard Humanoid Control https://megadose.ai/i/9639f7ff-9da8-42a9-8fed-72ec1e78fb5f Megadose en 2026-09-21T16:19:33+00:00 MedRSI: Recursive Self-Improvement for Medical Agents via Clinically Aligned Self-Evolution https://megadose.ai/i/df749f03-f9d5-468b-adc0-67a7ab19b298 Megadose en 2026-09-21T16:15:35+00:00 GRUET: Quantifying Uncertainty of Agentic Reasoning-and-Acting Processes https://megadose.ai/i/65d639af-d74c-42a5-9372-94c66265de56 Megadose en 2026-09-21T16:11:47+00:00 G-NAC: Graph Neural Automata Clustering via Emergent Domain Formation https://megadose.ai/i/a9d21929-97ab-4f30-ad1a-57162a2672d0 Megadose en 2026-09-21T16:11:01+00:00 The Answer-Basin Representation Hypothesis: We Are Not Probing or Steering Concepts https://megadose.ai/i/5844ff7d-ef6a-48f3-84d8-d19699b61851 Megadose en 2026-09-21T16:06:16+00:00 Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI https://megadose.ai/i/372b56da-24a9-4ce7-851a-766ba2f0efa0 Megadose en 2026-09-21T16:04:26+00:00 MSI-Bench: Evaluating Multi-Speaker Voice Interaction for Collaborative AI Agents https://megadose.ai/i/3fb4318e-081c-4f41-808f-7440a27f56ae Megadose en 2026-09-21T16:01:09+00:00 Decoding Guardrails: XAI-Guided Perturbation Analysis of Prompt Injection Detection https://megadose.ai/i/433fd166-c16f-4c60-909f-b1b3c2061d1d Megadose en 2026-09-21T15:59:06+00:00 When Quantization Preserves Accuracy but Not Evidence: Explanation-Aware Post-Training Quantization for Medical LLMs https://megadose.ai/i/9e30b410-dd44-42d0-b4ee-b0185e752211 Megadose en 2026-09-21T15:56:04+00:00 Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention https://megadose.ai/i/23939a99-01b5-4706-8889-8ba4339a78e0 Megadose en 2026-09-21T15:55:01+00:00 SpaceXAI releases Grok 4.7, which it says is better at verifying its own work and managing longer context, available for $2/1M input and $6/1M output tokens (xAI) https://megadose.ai/i/e173833b-3815-47c1-ab64-6aa2a2d501ee Megadose en 2026-09-21T15:50:15+00:00 Grok 4.7 https://megadose.ai/i/9948d042-e213-43a1-8f80-d0f9b7a042c0 Megadose en 2026-09-21T15:35:01+00:00 NYC-based Corridor, which uses AI to offer health benefits for small businesses, raised a $16M seed led by Bain Capital Ventures, following a $9M pre-seed (Brock E.W. Turner/Axios) https://megadose.ai/i/d21f8d44-6b9b-4ada-9ae3-604807cd6b6c Megadose en 2026-09-21T15:30:19+00:00 Improving synthesis prediction of small molecules at scale with RetroChimera https://megadose.ai/i/151fd4d2-2b5b-4da8-a8c8-df6b28c0b76c Megadose en 2026-09-21T15:15:00+00:00 Z.AI open sources its coding harness ZCode and disables certain features after users said ZCode was uploading codebases onto overseas servers without consent (Laurie Chen/Reuters) https://megadose.ai/i/984eab12-f16a-43bf-9327-e292fb5294a0 Megadose en 2026-09-21T14:39:22+00:00 Google’s $899 Googlebook is a bet that you’ll buy a new laptop for Gemini https://megadose.ai/i/4c166164-e7dc-414c-99e8-7f0a4efce8fc Megadose en 2026-09-21T13:45:00+00:00 Nscale's S-1: Microsoft and Anthropic account for 85% of its $103B in total contract value, only $2.6B of contract value was active as of late August, and more (Bloomberg) https://megadose.ai/i/ffd7b9f3-3c23-4db2-9c5b-141f85167179 Megadose en 2026-09-21T13:19:36+00:00 The Endless Exam: Mathematical Constructions from Today's Models toward Superintelligence https://megadose.ai/i/e4ad0e2e-9679-4d23-b11e-a4ce0ee84f75 Megadose en 2026-09-21T12:54:45+00:00 Not All Task Vectors Need Equal Rank: Energy-Proportional Allocation for Model Merging https://megadose.ai/i/a849b26e-d095-4de0-a72b-f9c7209aea57 Megadose en 2026-09-21T12:44:35+00:00 On Emergent Capabilities and Model Merging https://megadose.ai/i/c0f8caa1-79fb-4c2b-a8fd-f3a454c77277 Megadose en 2026-09-21T12:27:23+00:00 Lifted Bellman Linear Programming for Offline Reinforcement Learning https://megadose.ai/i/0b22e016-0355-4b9d-ad85-d9c931dde615 Megadose en 2026-09-21T12:26:13+00:00 AgentSTAR: Agentic Shape Tracking and Reconstruction from Monocular Videos https://megadose.ai/i/e5f725aa-0401-4bc7-bccf-4e5a02abf0bb Megadose en 2026-09-21T12:24:39+00:00 VPRune: Efficient Training-free Pre-LLM Visual Token Pruning https://megadose.ai/i/7d2ca5ab-9143-4ca8-84e3-a760bb89dc4c Megadose en 2026-09-21T12:19:50+00:00 Fathom-Vaidya: Advancing Medical Reasoning with Rubric-Based Rewards https://megadose.ai/i/0a32b08a-2fd9-46d3-ad91-c7db3b75af71 Megadose en 2026-09-21T12:08:13+00:00 RAILS: Retrieval-Augmented Incremental LLM Clustering at Scale https://megadose.ai/i/6d05d54e-3629-48e2-b151-8f320f8f2918 Megadose en 2026-09-21T12:02:41+00:00 Conduit: An Experience Data Plane for Distributed Reinforcement Learning https://megadose.ai/i/6dc1b4c6-b2b4-4cab-bbf4-776c8e0bde2f Megadose en 2026-09-21T11:59:07+00:00 Do LiDAR Language Models Really Understand Spatio-temporal Relationships? https://megadose.ai/i/621d27cf-3a6d-46d5-bd2e-525b842120cd Megadose en 2026-09-21T11:47:09+00:00 ActGov: Governing LLM Agent Actions via Policy-Constrained Validation https://megadose.ai/i/4412add8-d080-4536-a6bd-dbd518eedeab Megadose en 2026-09-21T11:35:50+00:00 Comparing Latent Concept Formation in State Space Models and Transformers via Sparse Autoencoders https://megadose.ai/i/dada770d-1252-4e1d-9db9-2c94b4d53915 Megadose en 2026-09-21T11:28:41+00:00 FoldQuantVLA: Native Low-Bit Quantization of Vision-Language-Action Models via Consistent Folding https://megadose.ai/i/8e21dd23-aee3-408b-b712-b4e44ce32b6f Megadose en 2026-09-21T11:28:24+00:00 1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation https://megadose.ai/i/14e760fa-0c37-4134-bb1f-80d46d418b5f Megadose en 2026-09-21T11:15:02+00:00 Prior-Amortized In-Context Bayesian Inference for Generalized Linear Mixed-Effects Models https://megadose.ai/i/bc61cae3-de09-4fd2-8574-960bf088aa8c Megadose en 2026-09-21T11:12:53+00:00 Finderchangchang/jev-chat-jarvis https://megadose.ai/i/83ba0fcd-9db4-4a8e-8693-bf16d4e31413 Megadose en 2026-09-21T11:12:53+00:00 jev-chat/jev-chat-jarvis https://megadose.ai/i/ef2f790f-87b9-4ffa-af6c-c083c717b974 Megadose en 2026-09-21T11:11:17+00:00 ARM: Attention with Routed-Memory for Learnable Sparse Control https://megadose.ai/i/f49372a5-15c9-41d5-af58-d95367bb9f42 Megadose en 2026-09-21T10:40:01+00:00 Sources: DeepSeek CEO Liang Wenfeng says training on Huawei chips is one of DeepSeek's biggest bets and Huawei is set to deliver training chips in Q4 or Q1 2027 (The Information) https://megadose.ai/i/f5f62b0a-4b78-4334-b2ae-41b581bf71e3 Megadose en 2026-09-21T10:22:06+00:00 Tactile-JEPA: Topology-Aware Self-Supervised Representation Learning for Distributed Tactile Sensors https://megadose.ai/i/da020558-f149-416a-a709-652eaac95929 Megadose en 2026-09-21T10:17:22+00:00 Information-Time Proximal Policy Optimization https://megadose.ai/i/7b46f4c6-d311-400d-b750-3808ad0b591a Megadose en 2026-09-21T10:15:51+00:00 Topographic Training Concentrates Causal Circuits Without Improving Neuron Monosemanticity https://megadose.ai/i/95b36bd3-0b18-4cd2-a187-909f37ed2cd6 Megadose en 2026-09-21T09:52:48+00:00 VLM-in-Sandbox: Visual Workspaces for Agentic Visual Reasoning https://megadose.ai/i/b2dc32ef-6df2-4ed8-a677-f3fa6a0efe5f Megadose en 2026-09-21T09:50:44+00:00 Sources: Alibaba names senior AI researcher Dayiheng Liu as the head of its Qwen project, bringing clarity after multiple reorganization rounds earlier in 2026 (Juro Osawa/The Information) https://megadose.ai/i/7863d1cf-0fa7-4ec5-a523-3c3e32f298bd Megadose en 2026-09-21T09:44:24+00:00 Few-Shot Demonstrations Elicit the Use of In-Context World Representations in LLMs https://megadose.ai/i/09879031-f9a7-4985-a600-ecc4d55c885a Megadose en 2026-09-21T08:45:15+00:00 Anthropic tests Fable 5.2 and Opus 5.5 ahead of the release https://megadose.ai/i/8a1fbc64-bdd4-4770-9899-5fb59aa0a021 Megadose en 2026-09-21T08:45:01+00:00 Big Tech is increasingly using residual value guarantees for off-balance sheet AI spend, giving Nvidia and Broadcom a way to support their customers' purchases (Financial Times) https://megadose.ai/i/a1c7d106-76b1-472d-8c77-1e4b6a786723 Megadose en 2026-09-21T05:55:00+00:00 A researcher used GPT-6 Astra to decipher a WWI German radio transmission from 1918, one of the 50 famous unsolved ciphers listed on a German science blog (prinz) https://megadose.ai/i/9f028c1d-ae79-49e8-abb0-c7ac74edbfb5 Megadose en 2026-09-21T04:05:32+00:00 Altis Labs, whose AI model analyzes CT scans taken during oncology trials and produces predictions tied to patient survival, has raised a $25M Series A (Natalie Breymeyer/Axios) https://megadose.ai/i/c120bfce-2b7d-4f48-a4ca-eb85d588c6bc Megadose en 2026-09-21T03:45:01+00:00 Sources: SoftBank is looking to issue $10B and €1B in debt for its OpenAI investment, in what would be one of the biggest junk bond sales ever (Bloomberg) https://megadose.ai/i/c42a35fc-a4be-42b0-8daa-496813c77edd Megadose en 2026-09-21T01:40:01+00:00 Mithrl, which integrates its proprietary biomedical world model with the R&D systems of pharmaceutical companies, raised a $20M Series A led by Obvious Ventures (Natalie Breymeyer/Axios) https://megadose.ai/i/7f5de2f8-43ee-48d8-9b79-c35d80982741 Megadose en 2026-09-21T01:00:44+00:00 UK-based Unit1, which develops "hyper-realistic" digital avatars of musicians to recreate classic concerts, raised nearly £15M from Balderton Capital and others (Alex Lawson/The Guardian) https://megadose.ai/i/a4b02ebd-8171-4b1f-b116-4f3d48108f65 Megadose en 2026-09-21T00:00:00+00:00 Document Retrieval-Aware Chunking (D-RAC): Universal Retrieval-Aware Ingestion of Enterprise Documents via PDF Normalization and Multimodal Markdown Conversion https://megadose.ai/i/7a679e5e-f384-4904-93ee-b3d015fe2a7e Megadose en 2026-09-21T00:00:00+00:00 CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies https://megadose.ai/i/0b6ed175-d0a7-4e05-bc24-db4750985a68 Megadose en 2026-09-21T00:00:00+00:00 1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation https://megadose.ai/i/7108d094-bbd3-43c1-83d1-8712655f9cfa Megadose en 2026-09-21T00:00:00+00:00 Harness-Zero: Harness Distillation via Agent-as-Harness https://megadose.ai/i/2fd1154c-078e-4572-b83d-ddf59224fef1 Megadose en 2026-09-21T00:00:00+00:00 Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents https://megadose.ai/i/26bc4f46-1350-44e1-8b52-ad454fea330a Megadose en 2026-09-21T00:00:00+00:00 VideoGen-Agent: Reinforcing Video Generation Agents https://megadose.ai/i/a09d88e2-9ff0-4ab3-84cb-00e4aaff0746 Megadose en 2026-09-21T00:00:00+00:00 Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies https://megadose.ai/i/e85ec066-0655-4b27-b2e6-f8132e497b21 Megadose en 2026-09-21T00:00:00+00:00 EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation https://megadose.ai/i/7da59280-5530-4e4b-a7d3-da46f12f1692 Megadose en 2026-09-21T00:00:00+00:00 Complex KDA: Understanding and Enhancing the Expressivity of Kimi Delta Attention https://megadose.ai/i/58ea8350-9fe3-49dc-9e3e-3d393bfb705e Megadose en 2026-09-21T00:00:00+00:00 Streaming Video Editing with Easy Adaptation https://megadose.ai/i/ebbf6c9e-0cb0-4455-8937-1310706c1fc2 Megadose en 2026-09-20T21:40:01+00:00 Alibaba releases Qwen-Image-2.1, a 7B open-weight model it says outperforms most closed-source models, with native transparency and up to ten reference images (Qwen)