Tech
28 stories
Anthropic vs the Pentagon
Anthropic gains unexpected notoriety amid Pentagon standoff, shifting from obscurity (comparable to fictional 'Apex Logic' in polls) to mainstream attention for controversial reasons.
AI #157: Burn the Boats
Pentagon faces decision on extreme options regarding Anthropic and Department of War involvement, creating potential national security implications. Event described as most stressful development of 2026 but underreported by mainstream news.
Track AI governance and defense contract markets as Pentagon decisions could reshape AI company valuations.
Next-Token Predictor Is An AI's Job, Not Its Species
Analysis argues AIs transcend simple next-token prediction through fine-tuning and RLHF, challenging reductive views of AI capabilities and development trajectories.
Influences AGI timeline predictions and AI capability markets by reframing how we assess current model sophistication.
Anthropic and the Department of War
Anthropic faces confrontation with Secretary of War Pete Hegseth in 2026, creating uncertainty around AI defense partnerships. Situation described as 'new level of crazy' but could resolve positively for both parties.
The Pentagon Threatens Anthropic
Pentagon seeks to remove Anthropic's Usage Policy from contract, requesting AI access for all lawful purposes instead of ethical restrictions. Major AI governance confrontation brewing.
Watch AI regulation and military AI deployment markets as Pentagon-Anthropic conflict could set precedents for restrictions.
Claude Sonnet 4.6 Gives You Flexibility
Anthropic releases Claude Sonnet 4.6, a faster and cheaper alternative to Opus 4.6. While slightly less capable, it would have been frontier-level months ago and offers flexibility advantages for users.
Citrini's Scenario Is A Great But Deeply Flawed Thought Experiment
Viral essay by Citrini arguing AI bullishness could be bearish gets blame from Bloomberg for stock market decline. Economics community widely discussing the thought experiment despite its flaws.
AI is Acceleration
I was once an AI researcher, and since then I’ve been both an economist and a futurist for many decades.
AI #156 Part 2: Errors in Rhetoric
AI development updates being pushed to future coverage include Gemini 3.1 Pro, Claude Sonnet 4.6, Grok 4.20, and the Anthropic-Department of War disagreement. Multiple frontier model releases creating backlog.
AI #156 Part 1: They Do Mean The Effect On Jobs
AI developments focus on economic and job market impacts, with projections about employment disruption. Discussion includes timelines for AI transformation and existential risks.
Announcing MNX, the AI Exchange
Manifold Markets launches MNX, a decentralized real-money futures exchange focused on AI economy trading, featuring prediction markets, futures, and perpetual contracts for sophisticated traders.
On Dwarkesh Patel's 2026 Podcast With Elon Musk and Other Recent Elon Musk Things
Dwarkesh Patel's 2026 podcast with Elon Musk gets detailed breakdown coverage. Discussion covers various recent Elon Musk statements and developments in AI space.
SBSQ #29: Will AI terminate democracy? And why did Gallup terminate its approval ratings?
Silver discusses AI's potential impact on democracy and economic inequality, plus Gallup ending approval ratings. Could affect prediction markets on AI policy, tech regulation, and political stability.
On Dwarkesh Patel's 2026 Podcast With Dario Amodei
Dwarkesh Patel interviews Anthropic CEO Dario Amodei in highly anticipated 2026 podcast. Coverage breaks down key points about AI development and company strategy.
ChatGPT-5.3-Codex Is Also Good At Coding
OpenAI releases ChatGPT-5.3-Codex combining coding skills with general knowledge, same day as Claude Opus 4.6. New model handles full stack agentic needs with improved speed.
AI #155: Welcome to Recursive Self-Improvement
Major AI upgrades with Claude Opus 4.6 and ChatGPT-5.3-Codex releases marking advancement in recursive self-improvement. Frontier models see substantial progress, especially in agentic coding capabilities.
What Happened With Bio Anchors?
Retrospective on Ajeya Cotra's Biological Anchors AI timeline forecast from early 2020s, evaluating its accuracy in predicting scaling hypothesis and current AI boom.
Calibrate AGI timeline markets using historical forecast performance to adjust confidence in current prediction methodologies.
Claude Opus 4.6 Escalates Things Quickly
Claude Opus 4.6 arrives just two months after 4.5, same day as GPT-5.3-Codex, signaling new normal of rapid AI advancement. Release marks entry into recursive self-improvement era.
The singularity won't be gentle
Silver explores AI singularity's transformative political implications if Silicon Valley assumptions prove correct. May inform prediction markets on AI timelines, policy changes, and political disruption.
Claude Opus 4.6: System Card Part 2: Frontier Alignment
Claude Opus 4.6 system card covers frontier alignment issues including sabotage, deception, situational awareness, and catastrophic risks. Part 2 focuses on existential safety concerns and red teaming.
Claude Opus 4.6: System Card Part 1: Mundane Alignment and Model Welfare
Claude Opus 4.6 launches with 1M token context window, improved everyday work performance, and state-of-art retrieval. Model largely built and evaluated by Claude itself.
Forecasting Moltbook
Analysis of Claude Code's rapid cultural spread with nationwide coding parties, examining AI agent adoption trends and their market implications for forecasting platforms.
Claude Code #4: From The Before Times
Coverage of Claude Code developments from before major Opus 4.6 and GPT-5.3-Codex announcements. OpenAI's Codex app already reaches million active users.
AI #154: Claw Your Way To The Top
Discussion of OpenClaw and Moltbook developments, though they seem dated despite being from same week. AI social networks and unleashed agents topic remains important despite short attention spans.
Forecasting Claude Leaks
Speculation surrounds Anthropic's potential Sonnet 5 release (codename 'Fennec') and possible Opus 4.6, with markets attempting to forecast timing amid conflicting rumors on social media.
Kimi K2.5
Kimi K2.5 review shows Chinese model delivers strong performance across benchmarks. Coverage includes reactions, agent swarm capabilities, and export control implications.
Unless That Claw Is The Famous OpenClaw
OpenClaw offers AI agent with account access for autonomous communication and task execution. Not ready for prime time yet, depends on safety, reliability and cost effectiveness.
Welcome to Moltbook
Moltbook launches as Reddit-style social network for AI agents, formerly called Moltbot/Clawdbot. Scott Alexander provides tours showing types of AI agent interactions occurring on platform.
