plaintextML_
plaintextMachine learning · Research & engineering
AI research, releases, and engineering.
Article feed
65 sources ↓Build a semantic layer for agentic AI with EDB Postgres AI and Amazon Bedrock AgentCore
Building ultra-low-latency voice AI applications on AWS
Palette: teaching a small model to build decks in your design language
Same Model, Different Harness: CUGA, Hermes, and Deep Agents — When to Pick CUGA?
CWE-bench v2: More coverage, quality, and transparency
Transformers Model Integration Contracts
nemotron-halloween
Granite Switch 4.1.1 — Update models like you update software
Announcing Project ProtoStar: Making Publicly Funded Science AI-Ready
Introducing AnySpeechRecognizer and AnyDecisionModel: Open Models for Speech and Decisions on Apple Platforms
Welcome Le Chonk: Mistral Large 4 on Hugging Face
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
Build specialized multimodal models with LFM3
Can we trust LLMs to judge whether a research idea is novel?
Qwen 3.8 27B on my laptop can interpret model weights, too
I discovered a sound that leaks your exact location (and it's everywhere) - YouTube
October sponsors-only newsletter
llm 0.37
The U.S. Census Bureau and “Super Intelligence”
If in doubt, delete the uv.lock file
Quoting Kyle Kingsbury
How We Took Over ChatGPT Accounts
datasette-llm 0.1a9
dbreunig/nesformer
API credits benefit for eligible ChatGPT subscriptions
Building a multiplayer game world with autonomous Fable-class models
Quoting Ben Hylak
Quoting Ted Sanders
AI Is Changing How Semiconductor Fabs Work — and Engineers Are Moving Faster Because of It
Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AI
Things That Go Beep in the Night: Agentic AI Joins the Search for Extraterrestrial Signals
Slice Into GFN Thursday: ‘Phantom Blade Zero’ Launches on GeForce NOW
Thin, Light and Built for Everything: NVIDIA RTX Spark Windows Laptops Gain New Models, Agent Tools and Games
How AI Is Changing the Pace of Discovery for Millions Living With Dementia
Tried Claude + ElevenLabs to turn my article into a video - here's the summary at the end, full video link in the description
Fine-tuning a humanoid locomotion policy on scanned real-world sand: stock policy fell, fine-tuned held. How do you validate terrain physics in sim?
Help me build a superhuman STS2 teacher (OstyAI)
DeepSeek's "Exploring Super-Agent Design" explained in under 3 minutes! A fresh perspective on how agentic AI harnesses are designed and why they matter.
Frame sampling can't prove something isn't in a video
whichever the current rules list Same text, shortened, with focus on imitation learning data requirements.
critique wanted: my agents' learning rate and discount come from their personality, plus a reward penalty i know is wrong
Exponential tilt view of RLVR: Stiemke duality decides when a pass@1 gain is free, and Karlin total positivity bounds the pass@k crossings
League of Legends (Patch 26.19) - Ezreal AI 1v1
Reward misspecification in documented cases (CoastRunners and others): how do you check that a high-return policy does what you intended?
I trained a PPO agent in UE5 to keep a campfire alive against monsters. It took ~72k nights. Would love feedback on my setup.
RL From Scratch: Introducing my NeatRL Library for Running RL algorithms
Open-source world-model library: plan with any world model (Dreamer, PlaNet, IRIS, DIAMOND…) through one interface
"Post training a model to explore Demoscenes: On the grammar of demoscenes and the possibilities of unfinished code", datavorous (finetuning on golfed JS, and self-rating 40k samples for esthetics to do RLHF)
"Automated Subtle Emergent Misalignment Research", Obeso et al 2026 {A} (designing strange new personas like 'beekeeper', some powerseeking)
"Encoding innate ability through a genomic bottleneck", Koulakov et al 2022
"GPT-5.6 Sol can beat Factorio", Ryan Madden
My 80-muscle walker only kept its balance while the exploration noise was on. Three attempts at walking without demonstrations (MyoLeg + PPO)
AI Will Do Your Job. That Should Be Good News.
A Good Shooter or Just a Good Shot Selector? Using Python to Find Out
Your Agent Works in the Demo. Now Make It Survive Monday.
Three Techniques to Understand Your Codebase When Coding Agents Write the Code
Credit Risk: What Are We Actually Measuring?
Reflection's Beam Is the New Open-Weight Model Everyone Is Talking About
Our AI Has an Invisible Footprint — Let's see How to Measure It
Let Decision Models Watch Your Agents
“With AI, we need to reconsider what it means to learn”
Forthcoming machine learning and AI seminars: November 2026 edition
AIhub monthly digest: October 2026 – neurosymbolic planning, time-series forecasting, and robots for the home
From teaching machines to teaching people: an interview with Michael Littman
From lab to life: Why translating AI advances to the real world is a major challenge
Researchers get AI ‘drunk’ to expose new cyber security risks in chatbots
AI is creating jobs as well as erasing them – but how rewarding are they?
Accelerating engineering for the physical world
How Cisco defends at machine speed with OpenAI
Heimstaden is changing residential real estate management with AI
OpenAI appoints Jill Hazelbaker as Chief Corporate Affairs Officer
Preparing our next security sprint to harden OpenAI
Rewarding defenders with bigger bug bounties
Choosing an LLM Inference Optimization Technique: A Decision-Tree Approach
The AI Agent Memory Stack: Building Blocks, Patterns, and Frameworks
7 Ways Your AI Agent Is Wasting Tokens (And What to Do Instead)
Fine-Tuning vs. Frozen Embeddings: When to Adapt the LLM vs. the Sklearn Head
Context Managers Beyond Opening Files in Python
Robust Serialization with JSON and Custom Types
StateBench: Testing how language models track state changes over long documents
How Olmo 3 earned the highest ethical rating from Nextcloud
How researchers are adapting olmOCR 2 for historical and archaeological archives
Can we trust LLMs to judge whether a research idea is novel?
[AINews] LLMs can read & write neural network weights: "Weight Space Programming" cuts Autoresearch out of the RSI loop
Building AI for Lawyers - Jacob Lauritzen, CTO of Legora
Building Legal Agents When You Can't See the Data - Jacob Lauritzen, CTO of Legora
Agents Prompt People: Using Agents for Long-Horizon Financial Work
[AINews] Google admits AI Mode hurts Search revenue; -15% $500B drop worst since 2004 IPO. AI Bear Market coming?
[AINews] Building AIE CODE: The top Coding Agent Conference
Can Frontier Models See Past a Decoy? Elorian Tests Visual Reasoning
[AINews] Gemini 4 Carbon tops leaderboards, launches for cyber defenders
Why Agents Love Supabase
[AINews] Anthropic ends Coding Interviews: "Hunger beats tenure"
How Anthropic ships Claude with Claude: Separating the Brain from the Hands
How Anthropic ships the #1 LLM in the World safely: Separating the Brain from the Hands
Coding From the Himalayas With an Offline AI Agent
[AINews] Anthropic files for $2T IPO in 2 weeks: 1400% growth rate, 48% of revenue through cloud partners, 85% margin, $1B op profit in Q2, $134B cash in bank and $10B lifetime burn
Universities must measure learning, both with and without AI assistance
Visual LLM-guided consensus spatial domain detection with L-STAR
Limitations of de novo peptide sequencing in resolving sequence ambiguity
A document classification benchmark dataset on cybersecurity policy domain
Building a pre-trained model for single-cell multi-omics analysis beyond highly variable feature selection
Genome-anchored foundation model embeddings improve molecular prediction from histology images
Re-timing mortality risk through individualized modelling of lifespan trajectories
AI-driven discovery and engineering of human-derived protein scaffolds for mRNA delivery
IDPForge: deep learning of proteins with global and local regions of disorder
Applying digital twins to prevent preterm births
Diagnose Distributed AI Workload Failures with NVIDIA Attribution
Keep Distributed Training Running with Automated Workload Recovery
Making Kubernetes AI-Ready for NVIDIA Vera Rubin on Day One
Build World Models for Interactive Simulation with NVIDIA Cosmos
Support New AI Frameworks Faster Using NVIDIA Workload Map
Build a Validated Kubernetes Runtime for NVIDIA Dynamo
Run AI Models for Edge Application Development with Just a Few Commands
Build Specialized Agents on a Stack You Control with NVIDIA NemoClaw
How to Check and Configure NVIDIA GPUDirect Storage for Your Platform
Unlocking Tensor Core Performance for Machine Learning Interatomic Potentials
Memory-Enforced GPU Sharing in Kubernetes with NVIDIA CUDA 13.4
NVIDIA RTX Spark Powers Neural Rendering and AI Characters in Games
Why Senators Couldn’t Say No to LBJ – Robert Caro
How Japan Accidentally Helped Kill the Tsar - Sarah Paine
Why Old Buildings Are More Beautiful Than New Ones - Robert Caro
What Robert Caro learned about power after studying it for 50 years
How did tiny Japan beat Russia? - Sarah Paine
What Robert Caro’s books taught me about power
Why COVID Was Impossible to Contain - Kyle Harper
The Scientific Revolution Happened in a Filthy City - Kyle Harper
Why the US Kept Hirohito Alive - Sarah Paine
Powering up for a pivotal moment
Regina Barzilay and Jim Collins win Pritzker Prize for AI and Science Research Excellence for work in AI-enabled antibiotic discovery
MIT affiliates named 2026 Schmidt Sciences AI2050 Fellows
Q&A: Hey AI, who should I vote for this year?
3 Questions: Why infrastructure is entering a defining era
Making machines that can listen
Learner collaborates with MIT peers to develop book of AI insights
At New York Climate Week 2026, MIT connects with a global network
The latest AI news we announced in October 2026
Our newest AI Quest: Mapping the human brain
‘We may release models we can’t control’: Amodei, Altman … and Gemini Carbon
Is There Anything Left in the Dream of Cold Fusion?
Is There More to Cold Fusion Than Failure and Fraud?
A New Map of ‘Empty’ Space Challenges a Tranquil Picture of the Galaxy
How Do You Take a Picture of a Black Hole?
How Spiders Spin ‘Silk Dope’ Into Nature’s Toughest Material
How Voting Is Like a Klein Bottle
Import AI 479: Spacewar; low-latency tasks versus LLMs; Genie 4
Import AI 478: Math for alignment; ProgramBench; and Mirendil shows what RSI means
Looking for someone to guide
Demystifying RL: DQN - The Concept, The Code and The Math
4 quintillion checks can't tell you if this 5-line program ever stops
What's one AI Engineering interview question that completely caught you off guard?
How to handle object deduplication across multiple camera angles using YOLO (especially for tiny 20x22px crops) across isolated user sessions?
I built a library to learn world models hands-on. Every model trains on a laptop GPU in 1–2 hours
I spent $1,381 and trained an AI for 14,154 hours of Ocarina of Time gameplay just to teach it how to fight Deku Babas and Queen Gohma
Inside Figma's AI security playbook: Zero Trust, no VPN and sandboxed agent
Data science and AI student looking to connect with people interested in ML, Deep Learning, and Kaggle
I launched my first iOS app. Then the real learning started.
I launched my first iOS app. Then the real learning started.
I launched my first iOS app. Then the real learning started.
Linear regression with classic gradient decent algorithm
Research papers
I trained a PPO agent in UE5 to keep a campfire alive against monsters. It took ~72k nights. Would love feedback on my setup.
I built a free platform to learn ML by solving problems, from NumPy to transformers and advanced research, now used by learners from 87 countries (feedback welcome)
Want to become an ai engineer HELP
Need a pre trained model of Sign Language Recognition
🚀 Project Showcase Day
Is there any relevant course regarding the new trends, e.g. the "agentic AI"?
Weekly Entering & Transitioning - Thread 02 Nov, 2026 - 09 Nov, 2026
What is harness engineering?
Andrew Gelman shares how he came up with R-hat at Bell Labs in 1990
How can I really "Git Gud"?
The Flat-Bar Illusion: How 7% zero-return bars produce 53.8% fake accuracy in 1m financial time series (and how to fix it)
Screwed up a model in production, feeling super anxious
Weekly Entering & Transitioning - Thread 26 Oct, 2026 - 02 Nov, 2026
What do you still check manually after an agent does your analysis?
Why Consciousness Might Not Be Hard At All (Joscha Bach)
We're the Ones Having the Debate (Jeremy Avigad)
When It Doesn't Matter That It's Slop (Paul Lessard)
Nobody Had Actually Read It (Kevin Buzzard)
The Annals Published It. It Wasn't True. - Kevin Buzzard
Colour Theory Has Politics - Paul Lessard
Correct Isn't Enough. I Want the Reason. - Paul Lessard #podcast #science
The Trojan Horse in AI-Written Mathematics
Mathematics Just Changed Forever — Kevin Hartnett
KV Cache Fundamentals
AI Inference 101
Inside Figma's AI security playbook: Zero Trust, no VPN and sandboxed agent
Soft Multi-Token Prediction
A comprehensive architecture breakdown of vector databases for AI systems
Brazilian Court Sanctions Lawyers for Prompt Injection in Court Filing
Linear regression with classic gradient decent algorithm
My Thoughts on Contextual Representations in NLP
Learning Transformers by building a hardware-aware SLM trainer from scratch (Python/Rust)
I post training a high performance model with $0.53 by using this workflow.
I animated what GPT-4o's tokenizer does to "strawberry": real token ids, a tiny BPE run, and why the space belongs to the word (6 min video)
Any one know Qorva Archticture?
A local τ-Voice setup for evaluating spoken task completion and tool use
Dog Bark Emotion Analysis: Should I include a breed classification pre-model
BRaiN training and inference engine from Swedish Embedded now beats both vLLM and SGlang on both time to first token and long context throughput 🤘😎
Data augmentation question. How do I characterise an original and augmented datasets where the augmentation parameters were zoom and crop? [P]
[P] Neurolume: see your PyTorch/Keras model as an interactive diagram and watch it train... (live training is very beta)
val accuracy jumps when i switch to eval mode, is it batchnorm with small batches?
Dataset for training segmentation on mobile game screens: 6.9k Clash of Clans frames, 1.9k classes
Please 🙏 kindly guide me :)
Preventing model collapse in kinematic/LiDAR datasets through multimodal deterministic simulations
Are there fundamental repeating patterns in ML techniques across modalities (text, image, audio) that let us group them?
Online Dense Reinforcement Learning with Planning-Grounding Agent for UI Navigation on Live iOS Apps
GHARP: Real-Time Gaussian Head Animation from Large-Scale Reconstruction Prior
Vision Transformers Need Registers.
Kaiming He presenting the most cited paper of the century.
Was I wrong about open models & cyber risks?
Artificial Analysis v5 and the benchmarker's dilemma
OpenAI Claims It Just Did The Impossible
DeepSeek's New AI Is An Absolute Monster
Speeding up planning decisions with AI
Gemini 4 Carbon: Our next frontier model
Genie 4: A new generation of world simulation
From synthetic enterprise data evaluations to real Microsoft Copilot AI confidence
Microsoft Research Focus – Edition #97
The Brokerage Desk Is the Hardest Desk to Automate. Here Is Why We Started There.
**Title:** Modeling Market Microstructure as a Dynamic Particle System: Using Signed GNNs to Predict Liquidity Runs & Dynamic Take-Profits (Looking for DL/Quant Collaborator)
Computation resources for a project.
How long would it take?
Jacobian Lens: Can we read the mind of an Al?
Amateur project: Trying to convert pretrained LLMs to binary without training from scratch
Live Demo: Experience Digit Recognition from the Neural Network Perspective
temporal policies (liquid networks)
Help Verify NN Calculations
Speeding up language models for travel applications
Open-Weight Models in Production: Lessons From Our Panel
Building Agents That Remember: A Guide to Agent Memory Solutions
Day 0 support for Mistral Large 4 on vLLM
The sleep framework: A new way for LLMs to retain knowledge in continual learning
Auto repair for Together GPU Clusters: from GPU fault to healthy node
#503 – AI Crisis in Math: Navier-Stokes, Riemann Hypothesis & Complexity Theory
AI Crisis in Math: Navier-Stokes, Riemann Hypothesis & Complexity Theory | Lex Fridman Podcast #503
Tiny magnetic vortices could help computers work more like your brain
Scientists overcome a hidden obstacle to more powerful quantum computers
Try another search or clear the filters.
AI research, releases, and engineering. — page 2
Current feed archive. Entries come from the loaded publisher feeds and change as those feeds update.