Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Minimalist editorial illustration in Anthropic style: a simple microchip icon with square center and radiating pins, thick black hand-drawn lines with slight wobble, centered on warm off-white background, flat vector shapes, subtle paper grain texture, lots of negative space, high contrast, 16:9 landscape with vertical split composition placing icon on left 60% panel.
The many eulogies for AI capability growth after the release of GPT-5 seem especially short-sighted right now. Letting people nervous about AI feel they can safely ignore AI development because it was pure hype that would never have any real impact is not a good thing for anyone”” https://x.com/emollick/status/2018141446683930811
AI leaderboard maker LMArena hits $1.7 billion valuation – Sherwood News https://sherwood.news/tech/ai-leaderboard-maker-lmarena-hits-usd1-7-billion-valuation/
Fueling the World’s Most Trusted AI Evaluation Platform https://arena.ai/blog/series-a/
LMArena has raised $150M+ at a valuation of $1.7B+ 💪🏼 In the past 7 months, @arena has: Grown our userbase 25x. 35M+ unique users. Grown our revenue from 0 to >>$30M+ ARR in 4 months. Our products help labs and enterprises measure the real utility of AI and understand their”” https://x.com/ml_angelopoulos/status/2008577473450250441
The industry is shifting from asking “What can this model do?” to “Can I trust it?” LMArena’s $150M raise underscores the growing need for independent, transparent, real-world evaluation frameworks that ensure AI systems meet the rigorous reliability and trust requirements of”” https://x.com/istoica05/status/2008575786169889132
Today, we’re excited to announce our $150M Series A at a $1.7B valuation–nearly 3× our May seed round. Since launching evaluations in Sept, our annualized consumption run rate has surpassed $30M. Our mission is clear: to measure and advance the frontier of AI for real-world use,”” https://x.com/arena/status/2008571061961703490
8 plots that explain the state of open models https://www.interconnects.ai/p/8-plots-that-explain-the-state-of
Jim Fan on X: “The Second Pre-training Paradigm” / X
https://x.com/DrJimFan/status/2018754323141054786
[2512.23236] KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta https://arxiv.org/abs/2512.23236
[2601.04171] Agentic Rubrics as Contextual Verifiers for SWE Agents https://arxiv.org/abs/2601.04171
Dynamic context discovery · Cursor https://cursor.com/blog/dynamic-context-discovery
AA-Omniscience is our benchmark measuring factual recall and hallucination across 6,000 questions covering 42 economically relevant topics within 6 domains (Business, Health, Law, Software Engineering, Humanities & Social Sciences, and Science/Engineering/Mathematics). Models”” https://x.com/ArtificialAnlys/status/2008570655047118914
New year, new Artificial Analysis Intelligence Index! Announcing Intelligence Index v4.0: incorporating 3 new evaluations, further aligning to real-word use and reducing saturation The Artificial Analysis Intelligence Index is our synthesis metric for assessing generalist model”” https://x.com/ArtificialAnlys/status/2008570646897573931
Shovel-ready short investigations seek funding! – How is AI’s adoption varying across roles, sectors, & regions? – Trends & bottlenecks for data supply? (incl. synthetic and RL environments) – Forecast for worldwide compute buildout? – Will inference costs continue falling? 🧵”” https://x.com/EpochAIResearch/status/2018767169094566208
Brendan Foody on Teaching AI and the Future of Knowledge Work (Ep. 267) | Conversations with Tyler https://conversationswithtyler.com/episodes/brendan-foody/
LLMs as Judges: Measuring Bias, Hinting Effects, and Tier Preferences https://aashidutt.substack.com/p/llms-as-judges-measuring-bias-hinting?triedRedirect=true
We cut VLM eval compute by >10× while INCREASING signal. The secret? Most benchmark samples are noise: → 70% solvable without the image → 42% mislabeled or ambiguous → MCQ formats hide 35-point capability gaps Presenting: DatBench 🧵 1/n”” https://x.com/HaoliYin/status/2008554232258113925
Lord of War, meet Lord of Tokens: Torture-testing image models on design-agency grade work | KAY SINGH https://singhkays.com/blog/lord-war-test-image-models/
Reinforcement Learning for Active Perception in Autonomous Navigation. [📍GitHub & Paper ] Most robots navigate as if their cameras were nailed in place. But perception is not passive. Animals move their heads and eyes constantly to decide where to go next. Robots should do”” https://x.com/IlirAliu_/status/2018762226170016109
A collection of Python robotics algorithms (localization, SLAM, path planning, motion planning). (📍GitHub ) Python sample codes and textbook for robotics algorithms. GitHub: https://t.co/4NRf4VCkZP —- Weekly robotics and AI insights. Subscribe free: https://x.com/IlirAliu_/status/2018404604556513504
Existential_Risk_and_Growth.pdf https://philiptrammell.com/static/Existential_Risk_and_Growth.pdf
[2512.24617] Dynamic Large Concept Models: Latent Reasoning in an Adaptive Semantic Space https://arxiv.org/abs/2512.24617
[2601.02346] Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling https://arxiv.org/abs/2601.02346
2026 Predictions: Much Faster Inference, Pre-Training with RL, and FP4 Everywhere https://kaitchup.substack.com/p/2026-predictions-much-faster-inference
Database Development with AI in 2026 – Brent Ozar Unlimited® https://www.brentozar.com/archive/2026/01/database-development-with-ai-in-2026/
Deep Delta Learning https://yifanzhang-pro.github.io/deep-delta-learning/
Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling | Falcon https://falcon-lm.github.io/blog/falcon-h1r-7b/
FinePDFs: Liberating 3T of the finest tokens from PDFs – a Hugging Face Space by HuggingFaceFW https://huggingface.co/spaces/HuggingFaceFW/FinePDFsBlog
Great data curation isn’t just for training! We @datologyai just released DatBench, a refined VLM eval suite with a simple motivation: VLM evals are broken. VLM evals are noisy, often measure the wrong thing, and expensive, often consuming ~20% of train compute. No longer!”” https://x.com/arimorcos/status/2008563285751476454
GRPO++: Tricks for Making RL Actually Work https://cameronrwolfe.substack.com/p/grpo-tricks
I liked some parts of the paper by @sarahookr https://t.co/GZcj8lvI4N but fundamentally it suffers from a problem I’ve seen repeatedly which is conflating how actual LLM researchers think about scaling with how the mass of semi-technical twitter onlookers think about scaling.”” https://x.com/_aidan_clark_/status/2008573653051642215
Increasing problem with publishing work on AI is that the publication process is much slower than working paper process, so when papers finally get full peer reviews, authors are asked to account for newer papers that are built on the paper under review! No real norms around this”” https://x.com/emollick/status/2018805872651276393
M2.1: Multilingual and Multi-Task Coding with Strong Generalization – MiniMax News https://www.minimaxi.com/news/m21-multilingual-and-multi-task-coding-with-strong-general
Mi:dm K 2.5 Pro scores -55 on the AA-Omniscience Index, driven primarily by a relatively high hallucination rate (92%)”” https://x.com/ArtificialAnlys/status/2008415408580178007
Recursive Language Models: the paradigm of 2026 https://www.primeintellect.ai/blog/rlm
The Second Pre-training Paradigm”” https://x.com/DrJimFan/status/2018754323141054786
Warp Specialization in Triton: Design and Roadmap – PyTorch https://pytorch.org/blog/warp-specialization-in-triton-design-and-roadmap/
What’s Ahead – Alien Processes, Domains, and Data Models https://practicaldatamodeling.substack.com/p/whats-ahead-alien-processes-domains
Whitepaper: Practitioner’s guide to reinforcement learning – Weights & Biases https://wandb.ai/site/resources/whitepapers/reinforcement-learning-ebook/
Eight Software Markets That AI Will Transform Differently https://davegriffith.substack.com/p/eight-software-markets-ai-that-will
A pretty bold commentary in Nature written by linguists, computer scientists and philosophers declaring “”by reasonable standards, including Turing’s own, we have artificial systems that are generally intelligent. The long-standing problem of creating AGI has been solved.”””” https://x.com/emollick/status/2018524111627325554
State of AI Data Connectivity Report: 2026 Outlook – CData Software https://www.cdata.com/lp/ai-data-connectivity-report-2026/
Unlock the secret to AI success | Forrester study https://miro.com/events/secret-to-ai-success-forrester-study/?src=-newsletter_glb
Leave a Reply