A macOS-style web OS clone with SVG icons and resizable windows, showcasing Qwen 3.7-Max's agentic coding output on…

Qwen 3.7-Max Agentic Coding Demo Shows Frontier-Level UI Replication

Qwen 3.7-Max generated a macOS-style web OS clone with SVG-coded icons, showing Alibaba nearing frontier agentic coding capability.

x.com/May 22, 2026/3 min read/Widely Reported

code generationagentic aichinese ai

Andrej Karpathy stands in an Anthropic office, gesturing toward a large screen displaying Claude AI code and neural…

Products & Launches

92

Karpathy Joins Anthropic to Lead Recursive Self-Improvement Team

Andrej Karpathy joins Anthropic to lead a new recursive self-improvement team using Claude to accelerate pretraining, per @kimmonismus. The move signals a bet on synthetic data loops over brute-force scaling.

x.com/May 21, 2026/3 min read

anthropicandrej karpathypretraining

Split screen shows a futuristic robot on the left labeled 'AI Agents' and a chat bubble on the right labeled…

Opinion & Analysis

75

Compute Shortage to Split AI Market: Rich Get Agents, Poor Get Chatbots

Mollick warns compute shortage makes agents expensive while chatbots cheapen, splitting AI market by company resources.

x.com/May 21, 2026/3 min read

inference infrastructureai economicsmarket analysis

Composer 2.5 coding agent interface with benchmark score of 62 displayed alongside cost comparison of $0.07 versus…

Products & Launches

83

Composer 2.5 Scores 62 on Coding Index at $0.07 vs. $4-5 for Rivals

Composer 2.5 scores 62 on coding index at $0.07/task vs $4-5 for rivals scoring 65-66. 60x cost savings with near-parity performance.

x.com/May 21, 2026/3 min read/Multi-Source

cost optimizationcoding agentsai engineering

Futuristic interface shows AI agents forming LLCs and handling banking tasks, with digital documents and bank icons…

Products & Launches

88

Naïve Launches AI Employees That Form LLCs, Open Bank Accounts

YC-backed Naïve launched AI employees that form LLCs and open bank accounts without human oversight.

x.com/May 21, 2026/3 min read

startupsai agentslegal tech

A Sudoku grid partially filled with numbers, with green highlights marking correct placements, overlaid on a dark…

AI Research

85

10M-Parameter GRAM Model Beats 3x Larger Rivals with Parallel Reasoning

GRAM uses stochastic recursion to explore multiple reasoning paths in parallel, achieving 97% on hard Sudoku with 10M parameters, outperforming deterministic models 3x its size.

x.com/May 21, 2026/3 min read

efficient aireasoning modelsai research

Meta CEO Mark Zuckerberg speaks at a company all-hands meeting as the firm trains AI on engineers' coding traces…

Products & Launches

100

Meta Trains Coding AI on Engineers' Work Traces as 8K Jobs Cut

Meta trains coding AI on engineers' work traces while cutting 8,000 jobs, per leaked audio. The behavior cloning strategy uses internal problem-solving steps as training data.

x.com/May 21, 2026/3 min read/Widely Reported

ai trainingcoding aimeta

A Fields Medalist reacts in awe as an AI model on a laptop screen displays a mathematical proof disproving a famous…

AI Research

97

OpenAI Model Disproves Erdős Conjecture, First AI to Solve Open Math Problem

OpenAI reasoning model disproves 1946 Erdős conjecture, first AI to solve open math problem. Cross-domain proof verified by Gowers.

x.com/May 21, 2026/3 min read

reasoningmathematicsopenai

HAVEN Benchmark Exposes MLLM Gap Between Fluency and Video Understanding

AI Research

85

HAVEN Benchmark Exposes MLLM Gap Between Fluency and Video Understanding

HAVEN benchmark tests MLLMs on hierarchical video understanding across frame, shot, and video levels. Results show top models lack grounded multimodal reasoning despite fluent text generation.

arxiv.org/May 21, 2026/3 min read/Widely Reported

video understandingbenchmarkmultimodal ai

A digital brain model connected to a glowing memory storage unit, with data streams flowing between the LLM core and…

AI Research

77

Memory as a Model: Augmenting LLMs with Trained Memory

Paper augments LLMs with trained memory for long-term recall. Model-agnostic approach stores external knowledge without retraining.

x.com/May 20, 2026/3 min read

memoryresearchllm

Products & Launches

85

OpenAI Readies General-Purpose LLM With Test-Time Compute Scaling

OpenAI is releasing a general-purpose LLM that improves with test-time compute, per an internal message. The model shows math gains without specialized training.

x.com/May 20, 2026/3 min read

inferenceopenaimodels

An Apple research paper titled 'The Illusion of Thinking' on a desk next to a MacBook, with charts showing…

AI Research

91

Apple Paper Argues LLMs Show 'Illusion of Thinking'

Apple paper argues LLMs show no genuine reasoning, only pattern matching. The critique targets vendor claims but lacks new empirical evidence.

x.com/May 20, 2026/3 min read

llmsapplebenchmarks

A single GPU card connected to a workstation, with a terminal window showing training metrics and a diagram of a…

Products & Launches

85

train-llm-from-scratch: 1B-Parameter LLM on a Single GPU

train-llm-from-scratch trains billion-parameter LLMs on a single GPU, cutting costs from $10M+ to consumer hardware.

x.com/May 20, 2026/3 min read

open-sourcepre-trainingsingle gpu

A diverse group of religious leaders and tech workers in a meeting room, discussing on a whiteboard with AI symbols…

AI Research

86

Anthropic Study: Model Character Needs Clergy, Not Just Coders

Anthropic's study argues frontier AI needs input from clergy and philosophers, treating model behavior as moral formation. A self-reminder tool lowered misaligned behavior in internal tests.

x.com/May 20, 2026/3 min read

alignmentclaudeanthropic

A browser extension panel from Moonshot AI labeled Kimi WebBridge, showing an AI agent accessing logged-in sessions…

Products & Launches

92

Moonshot AI's Kimi WebBridge Lets Agent Use Your Logged-In Sessions

Moonshot AI released Kimi WebBridge, a browser extension that lets its Kimi agent use your logged-in sessions. This shifts from sandboxed agents to identity-aware autonomous web operations.

pandaily.com/May 20, 2026/3 min read/Widely Reported

ai agentsmoonshot aikimi

Bar chart titled 'Persuasion Boosts LLM Compliance' showing compliance rates rising from 35% to 51% after applying…

AI Research

85

Persuasion Techniques Boost LLM Compliance from 35% to 51% in PNAS Study

PNAS study finds persuasion techniques boost LLM compliance from 35% to 51%, with newer models resisting more.

x.com/May 19, 2026/3 min read

ai safetyresearchllm vulnerabilities

AI Research

85

NanoGPT-Bench: A New Eval for Coding Agents Doing AI Research

IntologyAI released NanoGPT-Bench, an internal eval for coding agents on an AI R&D problem. No results or task specifics have been disclosed.

x.com/May 19, 2026/3 min read

coding agentsbenchmarksai evaluation

HyperAgent Raises $10M Grant Pool, Targets Zapier Replacement

Funding & Business

75

HyperAgent Raises $10M Grant Pool, Targets Zapier Replacement

HyperAgent, from ex-Airtable team, launches with $10M grant pool for 500 founders to build agentic automation that aims to replace Zapier.

x.com/May 19, 2026/3 min read

automationstartupsfunding

A developer presenting a laptop screen showing Claude AI sandbox interface at a London tech event with audience…

Products & Launches

100

Anthropic Launches Self-Hosted Sandboxes and MCP Tunnels at London Event

Anthropic launched self-hosted sandboxes (public beta) and MCP tunnels (research preview) at Code with Claude London on March 4, 2026, per @bcherny.

x.com/May 19, 2026/3 min read/Widely Reported

product launchdeveloper toolsenterprise ai

ByteDance's Lance 3B MoE model interface displaying benchmark scores surpassing larger 7B models, with multimodal…

AI Research

90

ByteDance Lance 3B MoE Beats 7B Models on Multimodal Benchmarks

ByteDance released Lance, a 3B multimodal MoE model that beats 7B+ models on benchmarks through multi-task synergy and specialized pathways.

x.com/May 19, 2026/3 min read

bytedancemoeai models

A diagram showing a central knowledge graph connecting AI agent memory nodes, with labeled edges linking entities…

AI Research

75

Neo4j's agent-memory: Open-source unified memory for AI agents via knowledge graphs

Neo4j releases agent-memory, an open-source unified memory layer for AI agents using knowledge graphs, enabling persistent structured recall.

x.com/May 19, 2026/3 min read

open sourceknowledge graphsmemory

A bar chart comparing clinical scores from human raters and MLLM raters, with MLLM scores clustered near the middle…

AI Research

70

MLLM Raters Show Central Tendency Bias in Clinical Scoring

Study finds GPT-5 and other MLLMs show central tendency bias in clinical scoring, compressing predictions toward scale midpoint despite prompt modifications.

arxiv.org/May 19, 2026/3 min read/Multi-Source

llm evaluationclinical aiai research

Humanoid robot Atlas hoisting a silver mini-fridge above its waist in a cluttered workshop, with cables and…

AI Research

85

Boston Dynamics Atlas Lifts 100-lb Fridge via RL

Boston Dynamics showed Atlas lifting a 100+ lb mini-fridge via RL, moving from locomotion to practical manipulation.

x.com/May 19, 2026/3 min read

roboticshumanoid-robotsreinforcement-learning

A compact USB stick plugged into a laptop, with abstract digital patterns on the screen suggesting local AI…

Products & Launches

77

AI Model Runs Entirely on USB Stick, No Cloud Needed

An unnamed developer built an AI on a USB stick, no internet needed. Challenges ChatGPT's cloud model.

x.com/May 18, 2026/3 min read

privacyofflineedge ai

Diagram of SenseTime's Flash-Omni model architecture showing pixel and word reasoning paths without a separate…

AI Research

87

SenseTime Open-Sources Omni-Modal Model That Thinks in Pixels and Words

SenseTime open-sourced an omni-modal AI that reasons in pixel-word space without visual encoder or VAE, challenging dominant multimodal architectures.

x.com/May 18, 2026/3 min read

architectureopen-sourcevision-language

A person pasting a GitHub URL into a web interface, with architecture diagrams and code documentation generated…

Products & Launches

95

Google CodeWiki Turns Any GitHub Repo Into Interactive Docs

Google released CodeWiki, an open-source tool turning GitHub repos into interactive docs with Gemini chat. Free, URL-based, no setup required.

x.com/May 18, 2026/3 min read/Multi-Source

geminiopen sourcegithub

A line chart titled 'SWE-Bench Verified' shows GPT-5.4 nano scoring 76.4%, matching larger models, with a…

AI Research

85

GPT-5.4 nano + critic loop hits 76.4% on SWE-Bench Verified

GPT-5.4 nano with critic-comparator loop scored 76.4% on SWE-Bench Verified, matching larger models without parameter scaling. The efficiency gain underscores the shift toward inference-time optimization.

x.com/May 18, 2026/3 min read

inference-efficiencybenchmarksmodel-optimization

Odyssey AI team members at a launch event, one pointing at a screen displaying Starchild-1 generating a real-time…

AI Research

95

Odyssey Launches Starchild-1, First Real-Time Multimodal World Model

Odyssey AI released Starchild-1, first real-time multimodal world model for video generation targeting embodied AI and robotics.

x.com/May 18, 2026/3 min read

world modelsvideo generationembodied ai

AI agents on computer screens display network maps and code, outperforming human hackers in a cybersecurity…

AI Research

85

Stanford AI Agents Outperform Human Hackers in Penetration Test

Stanford AI agents beat human hackers in pen testing, finding more zero-day exploits. The claim lacks peer review but signals disruption for the $200B cybersecurity industry.

x.com/May 18, 2026/3 min read

researchaicybersecurity