https://megadose.ai/i/d8648edd-cb9a-4923-ac69-99367de796d7 Megadose en 2026-08-05T03:10:00+00:00 Cloudflare announces Cloudflare Wallets for users to pay for APIs and content via stablecoins, with the ability to create virtual wallets for agentic shopping (Jeff John Roberts/Fortune) https://megadose.ai/i/4f94dd7c-15d8-4a2c-8886-c01ca88cfb30 Megadose en 2026-08-04T23:58:24+00:00 New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging https://megadose.ai/i/689709fc-f81f-4d94-b6a8-0fa58b06183e Megadose en 2026-08-04T22:21:08+00:00 Alibaba released Qwen3.8-Max with open weights coming soon https://megadose.ai/i/75550e45-0f2e-4af0-84d8-b8be54872a80 Megadose en 2026-08-04T21:15:04+00:00 SpaceX says its Q2 capex was $18.4B, up from $2.8B a year ago, including $15.8B for AI (Micah Maidenberg/Wall Street Journal) https://megadose.ai/i/e5401330-cb08-4e36-96e0-0cfc7d451bff Megadose en 2026-08-04T21:07:26+00:00 SpaceX has bought $329M worth of Tesla Megapacks so far this year https://megadose.ai/i/beda3bd8-c5c9-4d8a-a2f6-5c4c266a69ac Megadose en 2026-08-04T20:51:07+00:00 AMD reports Q2 revenue up 50% YoY to $11.5B, vs. $11.3B est., Data Center revenue up 107%, forecasts Q3 revenue below some estimates; AMD drops 7%+ after hours (Ian King/Bloomberg) https://megadose.ai/i/b33899ab-f331-42d9-bc5c-89e36987d737 Megadose en 2026-08-04T20:25:00+00:00 Samsung previews zHBM, which vertically stacks HBM atop AI accelerators, and zNAND-O, built on V-NAND, and unveils the industry's first V10 BV-NAND architecture (Yoolim Lee/Bloomberg) https://megadose.ai/i/3bf21faa-da86-4ad2-b937-89a9d2d37587 Megadose en 2026-08-04T19:48:40+00:00 Anthropic signs $10 billion deal with AI cloud startup Volta https://megadose.ai/i/a6a62fb8-3a90-41cd-bcbe-f1c395bd3149 Megadose en 2026-08-04T19:34:34+00:00 Meet Wrinkles, an AI app that uncovers the hidden stories of the places around you https://megadose.ai/i/93a390f7-cb1d-4fe0-bbb8-0975dc0886db Megadose en 2026-08-04T18:55:01+00:00 Mistral releases Shieldstral, a 3B multimodal safety classifier that it says matches models up to 7x its size on text safety, available under Apache 2.0 (Mistral AI Blog) https://megadose.ai/i/d75a015d-f0af-4057-8ceb-644310632c35 Megadose en 2026-08-04T17:59:58+00:00 ParVL: Parallel Scaling and Expandable Compute Allocation for Multimodal LLMs https://megadose.ai/i/e0b6481f-0803-4128-8f92-71c6c8d4fc00 Megadose en 2026-08-04T17:59:56+00:00 SocietyBench: Forecasting Counterfactual Social-World Evolution https://megadose.ai/i/f6b871ba-251e-4734-a5b2-51450190b6c3 Megadose en 2026-08-04T17:59:55+00:00 WorldCup Arena: Prospective, Leakage-Free Evaluation of Frontier LLMs on a Live Tournament https://megadose.ai/i/5b2faf0c-fadd-42ac-b7a2-52a4cd6bde0d Megadose en 2026-08-04T17:59:21+00:00 TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning https://megadose.ai/i/83a1907b-a5ef-4af9-a5ac-b59e2426f93b Megadose en 2026-08-04T17:57:20+00:00 Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility https://megadose.ai/i/78ae0716-8520-439d-a656-94d108ed34cd Megadose en 2026-08-04T17:56:49+00:00 Agogic: Performance-Timed Music Tokens for LLM-Native Text-to-Symbolic-Music Generation https://megadose.ai/i/b4c10d56-27a9-49f1-847a-fd42475f8ff2 Megadose en 2026-08-04T17:54:01+00:00 When Attention Goes Blind: Numerical Failure in ALiBi Positional Encodings https://megadose.ai/i/468024d2-f39e-40ba-947b-919373bb5cf9 Megadose en 2026-08-04T17:47:25+00:00 Can Large Language Models Recover Semantic Optimization Opportunities That Compilers Miss? https://megadose.ai/i/f0eea632-0ab3-4141-a4bb-3ece0dcb774e Megadose en 2026-08-04T17:40:08+00:00 ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning https://megadose.ai/i/d2c9fc8f-e419-4d45-a39e-165dede3ebfc Megadose en 2026-08-04T17:38:06+00:00 Should We Type or Talk to LLM Agents? A Comprehensive Study of Voice and Keyboard Input Perturbations https://megadose.ai/i/ff7f25c8-7811-4a92-9b6c-cb92dc1f7866 Megadose en 2026-08-04T17:34:14+00:00 Information-Geometric Forward Policy Training in GFlowNets https://megadose.ai/i/19733138-dc66-42c5-bf17-447aa8845320 Megadose en 2026-08-04T17:28:24+00:00 Separating quantum circuits from classical LLMs https://megadose.ai/i/d71b5216-44b6-4e2b-a46d-fa0230e05cf7 Megadose en 2026-08-04T17:27:21+00:00 Interpretable Adaptive Sampling for LLM Test-Time Scaling https://megadose.ai/i/365cceb6-1d96-4a84-a7f4-4f055aa83792 Megadose en 2026-08-04T17:15:01+00:00 Anthropic names Mariano-Florentino Cuéllar, an ex-California Supreme Court justice and a special assistant in Obama's WH, as its first global affairs chief (Jeffrey Dastin/Reuters) https://megadose.ai/i/73d6ebd5-37cb-49eb-8073-8544c917b6f9 Megadose en 2026-08-04T17:10:47+00:00 Muon Meets Mamba: Spectral Optimization for State Space Models https://megadose.ai/i/c7644f89-b090-49be-b627-1cf2b3215962 Megadose en 2026-08-04T17:02:34+00:00 Logic Before Language: Pre-pretraining on Formal Derivations Fosters Skill Acquisition and Compressibility https://megadose.ai/i/f0ea702f-d7da-4a35-8438-bc91a36db43e Megadose en 2026-08-04T17:00:52+00:00 Latent Reward Registers for Diffusion Preference Alignment https://megadose.ai/i/56441a37-11f6-4b29-bd10-a5a7b6795912 Megadose en 2026-08-04T17:00:50+00:00 Sandisk and SK Hynix announce the High Bandwidth Flash (HBF) open specification that enables up to 512GB memory modules with 0.4 TB/s to 3.0 TB/s bandwidth (Anton Shilov/Tom's Hardware) https://megadose.ai/i/96a56b0b-b663-4403-8b2a-9862f07020e9 Megadose en 2026-08-04T16:49:53+00:00 When and Where to Look: Adaptive Visual Evidence Scheduling for Efficient Long Video Understanding https://megadose.ai/i/f684af48-1857-4cce-a663-dba1795d8faa Megadose en 2026-08-04T16:42:04+00:00 Trajectory inference via Acceleration Matching https://megadose.ai/i/e74cdedc-c694-4d50-a10e-2cab7f355db5 Megadose en 2026-08-04T16:40:48+00:00 Sparse Weight Decomposition for Efficient Circuit Extraction https://megadose.ai/i/254903ab-6e12-45dd-846b-ca3b1edd02f3 Megadose en 2026-08-04T16:26:47+00:00 Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse https://megadose.ai/i/c64b2260-305a-460b-84e7-92130784ab4f Megadose en 2026-08-04T16:25:38+00:00 Intertemporal Preference Steering in Qwen3 via Contrastive Activation Addition https://megadose.ai/i/986fff24-ac39-4aa1-8ba6-7d4841869552 Megadose en 2026-08-04T16:23:39+00:00 CARE-X: Towards Clinically Useful Radiology VLMs with Auxiliary Supervision, Reward-Aligned Learning, and Tool-Augmented Measurement https://megadose.ai/i/ea83935d-d830-4603-9c0f-5e9a724e7605 Megadose en 2026-08-04T16:22:30+00:00 Omega-S: A Functional Resilience Index for LLM Fine-Tuning https://megadose.ai/i/394ebade-2c93-4575-ad61-867aa36a2f3f Megadose en 2026-08-04T16:20:53+00:00 BanglaWild: An In-the-Wild Bengali Scene Text Recognition Benchmark for OCR and Vision-Language Models https://megadose.ai/i/d9880e48-0678-4056-ada4-3c9b8de22058 Megadose en 2026-08-04T16:18:52+00:00 MultiGlobeQA: A Multilingual and Globally Diverse Benchmark for Geospatial Reasoning https://megadose.ai/i/020800b8-1387-4fe5-83e1-ee5ca97c889e Megadose en 2026-08-04T16:15:12+00:00 Enhancing VLM Reward Models Through Structure-Aware Fine-Tuning https://megadose.ai/i/f50d47f9-64b5-4d3e-90f8-d38bf17b3960 Megadose en 2026-08-04T16:15:02+00:00 ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities? https://megadose.ai/i/67b5d3c5-018c-4510-8227-d8de85f385ea Megadose en 2026-08-04T16:10:39+00:00 When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation https://megadose.ai/i/9412f55d-b84a-46de-b1ad-a6e0f54c92d2 Megadose en 2026-08-04T16:09:01+00:00 GENESIS: Towards Explainable Causal Discovery https://megadose.ai/i/cd7a2a06-3e57-4963-9500-cc692d73c6a0 Megadose en 2026-08-04T16:03:45+00:00 Beyond Representational Similarity: Source-Conditioned Description-Length Gain for Generative Plagiarism Detection and Candidate Source Reranking https://megadose.ai/i/34a4c724-478f-4cc0-83ac-16656946bedd Megadose en 2026-08-04T15:50:06+00:00 Spotify expands AI remix and covers project with Merlin partnership https://megadose.ai/i/0e4c0d3b-8da8-4e6a-b714-9c43d02daac3 Megadose en 2026-08-04T15:35:02+00:00 Alpamayo 2 Super, Nvidia's frontier open reasoning model for robotaxis and AVs, is now available for commercial use under the OpenMDW-1.1 license (Jessica Soares/NVIDIA) https://megadose.ai/i/cfef660a-4274-4ea2-ae31-b0b0d826dc34 Megadose en 2026-08-04T14:20:01+00:00 Convex, which provides an AI-optimized application backend for developers, raised a $57M Series B led by Insight Partners, taking its total funding to $110.5M (Maria Deutscher/SiliconANGLE) https://megadose.ai/i/7ee51f65-425d-4fe6-844e-361b2039da57 Megadose en 2026-08-04T13:15:01+00:00 Obsidian Security, which helps companies secure AI agents, raised an $85M Series D at a $1.1B valuation, after raising a $90M Series C in April 2022 (Akash Sriram/Reuters) https://megadose.ai/i/ad9122d6-643a-41a3-a13c-d54f2cb1299c Megadose en 2026-08-04T13:05:01+00:00 HappyRobot, which uses AI to automate freight operator communications, raised a $150M Series C at a $1.2B valuation, after a $44M Series B in September 2025 (Lily Mae Lazarus/Fortune) https://megadose.ai/i/eebad75b-78a7-42aa-9a10-baa557e89ed5 Megadose en 2026-08-04T12:25:29+00:00 Sources: Anthropic agreed to a $10B deal for computing capacity in Norway from Nvidia-backed AI cloud startup Volta; Volta says the deal runs for six years (Bloomberg) https://megadose.ai/i/dc04aa0b-f83c-4fe5-98af-a42f041c8a14 Megadose en 2026-08-04T12:00:26+00:00 Introducing Shieldstral. https://megadose.ai/i/09998c06-110b-4f73-a60b-27cb498e4201 Megadose en 2026-08-04T12:00:00+00:00 Tino Cuellar joins Anthropic as Chief Global Affairs Officer https://megadose.ai/i/f82e02f9-cdff-4b11-a982-742256a0f59f Megadose en 2026-08-04T11:35:03+00:00 AI cloud startup Volta raised $300M co-led by a16z and Altimeter at a $2.4B valuation and says it landed a $10B contract with an unnamed leading AI developer (Dina Bass/Bloomberg) https://megadose.ai/i/88e303c6-a05c-4019-99fc-533015221cd4 Megadose en 2026-08-04T10:07:06+00:00 Why Large Language Models Fail at Tabular Prediction https://megadose.ai/i/4dc8e60e-8645-416c-b968-9d22282889cd Megadose en 2026-08-04T08:25:00+00:00 Sources and filings: Google assembled a ~$200B financing program for Anthropic, with $150B+ tied to TPUs and involving Broadcom, Blackstone, Apollo, and others (Financial Times) https://megadose.ai/i/90d4f95a-6c7d-43b2-b9f9-5d5826fb166d Megadose en 2026-08-04T03:49:14+00:00 [AINews] Qwen 3.8 Max(2.4T) and 27B, new open weights models for Coding and Cowork https://megadose.ai/i/227f64a1-c4d7-4a4d-9c79-a1513277c0ce Megadose en 2026-08-04T00:00:00+00:00 New ways to learn and teach with ChatGPT Work and Codex https://megadose.ai/i/492eeb81-f0f4-4ec0-931a-857242427682 Megadose en 2026-08-04T00:00:00+00:00 Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent https://megadose.ai/i/cf2047a4-6c19-4c78-8f2b-25964d751388 Megadose en 2026-08-04T00:00:00+00:00 PAST-Bench: Benchmarking the Foundations of Recursive Self-Improvement in Personal Agents https://megadose.ai/i/eb7b985e-dbc4-4593-a112-924b673f4313 Megadose en 2026-08-04T00:00:00+00:00 OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models https://megadose.ai/i/3098dfd3-5c94-4423-9e47-1ff4a9f29f2b Megadose en 2026-08-04T00:00:00+00:00 JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion https://megadose.ai/i/e249e196-70c6-478d-8f53-e58e79d10e3c Megadose en 2026-08-04T00:00:00+00:00 SkillJack: Persistent Skill Backdoors in Self-Evolving Agents https://megadose.ai/i/b434e467-269e-4075-a7e4-6e1b4b32e982 Megadose en 2026-08-04T00:00:00+00:00 UniWorld-Design: From Pixel Generation to Layer-Native Design https://megadose.ai/i/c8e77e95-29af-492d-be73-3b6d4cf9e5d8 Megadose en 2026-08-04T00:00:00+00:00 Any-OPD: Heterogeneous On-Policy Distillation for Flow-Matching Models via Representation-Space Bridging https://megadose.ai/i/c238689c-1e1a-46a8-8cc6-5bf7d27308a7 Megadose en 2026-08-03T20:35:01+00:00 Palantir reports Q2 revenue up 93% YoY to $1.94B, vs. $1.8B est., US commercial revenue up 149%, and raises FY 2026 revenue guidance; PLTR jumps 9%+ after hours (Samantha Subin/CNBC) https://megadose.ai/i/15980416-d830-4c1e-9697-4b691c72123d Megadose en 2026-08-03T20:00:00+00:00 AWS is helping vibe-coding startup Superblocks, and the implications are big https://megadose.ai/i/0d441a14-2485-4c53-9e6f-3d5dae243b2f Megadose en 2026-08-03T18:43:43+00:00 Apple finally fixed Siri. So why does it feel anticlimactic? https://megadose.ai/i/7f005d56-43cf-4ff9-a694-77cd16fc83d0 Megadose en 2026-08-03T17:59:50+00:00 AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling https://megadose.ai/i/5fe97bd8-8580-4564-b727-5c6ad547faba Megadose en 2026-08-03T17:58:27+00:00 onepot-Bench 0: towards lab-aware in silico chemistry benchmarks https://megadose.ai/i/a0535903-251f-4702-b5a7-85540b89d64a Megadose en 2026-08-03T17:55:24+00:00 GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning https://megadose.ai/i/728298cf-7e37-4257-9fac-4e4a6f55fee1 Megadose en 2026-08-03T17:51:58+00:00 CoWAM: Coordination Contracts for Selective Policy Intervention with WAMs https://megadose.ai/i/66da38b1-7c9d-4b99-a721-25858881215a Megadose en 2026-08-03T17:50:54+00:00 Pseudorandom Streams within Diffusion Models Act as Learnable Inputs That Affect Generation Quality https://megadose.ai/i/cdfe0a85-f78f-4474-a095-71526d90ee27 Megadose en 2026-08-03T17:45:58+00:00 AtumAI: A Principled Framework for Agentic Generation of Datacenter Control-Plane Policies https://megadose.ai/i/697a7c1a-090d-4c96-8b79-11039769fd0b Megadose en 2026-08-03T17:43:36+00:00 Structured Memory for Edge Language Models: Persistent Context and Corpus Retrieval via O(1) SSM State Injection https://megadose.ai/i/4cecd630-159c-4387-9ebf-01c1e7ca7c68 Megadose en 2026-08-03T17:41:20+00:00 Benchmarking Sheaf Neural Networks for Inductive Tasks https://megadose.ai/i/43b9a5c6-36a1-4293-92a5-16e317a44b8e Megadose en 2026-08-03T17:24:27+00:00 Uncertainty Is Not Enough: Value-of-Information Routing for Mixtures of LoRA Experts https://megadose.ai/i/62a823fb-892b-4ae2-8944-658361116aa9 Megadose en 2026-08-03T17:12:05+00:00 LiveMem: Maintaining Memory State Continuity in Long-Running LLM Inference https://megadose.ai/i/49b75e6e-0314-4c56-a514-09f940d02cb8 Megadose en 2026-08-03T17:07:50+00:00 RoMeRL: Balancing Feedback Coverage and the Memory-Reward Trap in Self-Evolving Agent Memory via Reduced-Order Utility States https://megadose.ai/i/654bbc03-5cd0-4b2f-8ef2-2e047cddd890 Megadose en 2026-08-03T17:04:43+00:00 CMuon: Accelerating and Stabilizing Diffusion Transformer Training via Chunked Momentum Orthogonalization https://megadose.ai/i/e258508b-6e3d-413f-81d6-a36d7df3d8e1 Megadose en 2026-08-03T17:03:19+00:00 SWE-Touch: Benchmarking Coding Agents When Users Touch the Code https://megadose.ai/i/7475b083-097b-4a99-996b-f67c40ae4955 Megadose en 2026-08-03T16:53:29+00:00 Computational and Statistical Guarantees of the \textit{c}-Rectified flow https://megadose.ai/i/fe03ac3d-df3f-485c-b420-02d6cc5ce333 Megadose en 2026-08-03T16:34:46+00:00 Real-Time Detection and Repair of LLM Agent Failures https://megadose.ai/i/818f0ba3-5286-4f48-8cf4-56e3552b21f5 Megadose en 2026-08-03T16:22:51+00:00 ParEvalLayer: When Partial LLM-Agent Evaluations Support a Decision https://megadose.ai/i/4da7b4b1-1f4d-4087-89b8-b51ed8117a26 Megadose en 2026-08-03T16:21:56+00:00 Right Answer, Wrong Method: Shortcut Hacking Misleads the Evaluation of LLM Reasoning on Frontier Science Benchmarks https://megadose.ai/i/7a45b973-20af-4888-a9f7-145e9f283373 Megadose en 2026-08-03T16:20:17+00:00 Agentic Commerce World: An Auditable and Verifiable Environment for Vibe Commerce https://megadose.ai/i/a629a7b0-551f-488c-bc52-6164eda04c00 Megadose en 2026-08-03T16:18:22+00:00 xPress: Parallel Refinement for Diffusion Drafters in Speculative Decoding https://megadose.ai/i/f879bfef-2b23-453a-a320-e37d3a47b3bd Megadose en 2026-08-03T16:15:01+00:00 Valar, which is making small modular nuclear reactors to power data centers, raised a $1B Series B led by Sequoia at a $6B post-money valuation (Bloomberg) https://megadose.ai/i/c4e82a1e-5e2a-4f2b-b645-5671c659da80 Megadose en 2026-08-03T16:10:29+00:00 Wasserstein mixing time of the unadjusted Langevin algorithm https://megadose.ai/i/f3e4e072-b882-4970-a8d0-3ddc0621aba1 Megadose en 2026-08-03T16:10:02+00:00 Mariana Minerals, which builds and operates mines using its MarianaOS software platform, raised a $310M Series B, bringing its total funding to $400M (Allie Garfinkle/Fortune) https://megadose.ai/i/10769ef0-979c-4460-b469-af11326d490e Megadose en 2026-08-03T16:03:16+00:00 Agentic Incident Response through Digital Twin-Enhanced Multiscale Planning https://megadose.ai/i/865bf9f5-0a80-4c53-a353-ee5b997d5950 Megadose en 2026-08-03T16:00:00+00:00 Orchard: An open framework for scalable agentic AI https://megadose.ai/i/2e41e18b-5b13-421f-8320-bb09bd0262ed Megadose en 2026-08-03T13:46:04+00:00 Zenity, which develops a platform for securing AI agents, raised a $125M Series C led by Norwest Venture Partners, taking its total funding to ~$185M (Meir Orbach/CTech) https://megadose.ai/i/7a10bf97-48e5-4b4e-b87b-8b4616d7f196 Megadose en 2026-08-03T13:16:04+00:00 Visa plans to acquire Israeli company BioCatch, which uses AI to detect fraud and distinguish legitimate users from attackers in real time, for $2.4B in cash (Katherine Hamilton/Wall Street Journal) https://megadose.ai/i/8101df0b-bdaa-4df8-ac4d-7d5b97a8bcdc Megadose en 2026-08-03T13:01:58+00:00 Horizon3, whose AI penetration testing platform NodeZero helps find and exploit attack paths in production systems, raised a $250M Series E at a $2B valuation (Kate Park/TechCrunch) https://megadose.ai/i/f5cb956a-e4bc-400d-ac96-858e43c49853 Megadose en 2026-08-03T11:50:00+00:00 June, which aims to help enterprise AI deployment by finding bottlenecks and building agents, emerges from stealth with $20M led by Marc Benioff's Time Ventures (Tim Fernholz/TechCrunch) https://megadose.ai/i/01ccb6e8-8702-4d07-b0b2-775d695dc878 Megadose en 2026-08-03T11:15:48+00:00 AirLLM 70B inference with single 4GB GPU https://megadose.ai/i/407f38b5-47b9-4ff8-a4f9-6248046b67ab Megadose en 2026-08-03T11:01:36+00:00 Sources: South Korean AI chip designer DeepX raised a ~$29M Series D at a ~$2.2B valuation, and is in talks to raise ~$209M at a ~$2.4B valuation by September (Yoolim Lee/Bloomberg) https://megadose.ai/i/c6d1c094-444a-4360-8078-51054264dd20 Megadose en 2026-08-03T10:45:58+00:00 Alibaba prices Qwen3.8-Max at $2 per 1M input tokens and $6 per 1M output tokens via its API, below Kimi K3's $3/1M input tokens and $15/1M output tokens (Henry Siu/The Information) https://megadose.ai/i/7dbd9e3c-a9c9-4efd-b562-2023d810d589 Megadose en 2026-08-03T10:30:55+00:00 Artificial Analysis: DeepSeek's V4-Flash costs $0.14/1M input and $0.28/1M output tokens, or $0.03 per test, far below Kimi K3's $0.86 and GPT-5.6 Sol's $1.86 (Eduardo Baptista/Reuters) https://megadose.ai/i/ae9cda75-8f5c-4896-9b4d-135d2e8a4cb0 Megadose en 2026-08-03T10:00:00+00:00 A Marc Benioff-backed startup thinks AI can solve the AI deployment problem