Author: Santiago Maniches
-
How Claude Solved a Problem Donald Knuth Could Not: The Math Behind “Claude’s Cycles”
Donald Knuth published ‘Claude’s Cycles’ on February 28, 2026, crediting Claude Opus 4.6 with solving an open graph theory conjecture he had been stuck on for weeks. The…
-
Harvey Hits $11 Billion: What Legal AI’s Fastest-Growing Company Reveals About the Application Layer
Legal AI startup Harvey raised $200 million at an $11 billion valuation on March 25, 2026, jumping $3B in three months. With 100,000 lawyers on the platform, $190M…
-
Langflow RCE Exploited in 20 Hours: How a Single API Endpoint Gave Attackers the Keys to AI Pipelines
CVE-2026-33017 (CVSS 9.3) is an unauthenticated RCE in Langflow exploited within 20 hours of disclosure with no public PoC. Attackers harvested OpenAI, Anthropic, and AWS API keys from…
-
ARC-AGI-3 Drops Frontier AI Models Below 1%: The First Benchmark That Tests Whether AI Can Actually Learn
ARC-AGI-3 launched March 25, 2026 as the first interactive AI benchmark. Every frontier model scored below 1% (Gemini 0.37%, GPT-5.4 0.26%, Claude 0.25%, Grok 0.00%). Humans scored 100%.…
-
Qwen 3.5 9B Matches Models 13x Its Size: What Small Models Mean for Edge AI
Alibaba released Qwen 3.5 9B on March 2, 2026: a 9-billion-parameter model that outperforms OpenAI’s GPT-OSS-120B (13x larger) on GPQA Diamond, MMLU-Pro, and multilingual benchmarks. The hybrid Gated…
-
Apple’s AI Reckoning: Why Siri Runs on Google’s Gemini Now
Apple and Google announced a multi-year deal on January 12, 2026 to power Siri with Gemini models, reportedly worth $1 billion per year. Apple tested OpenAI and Anthropic…
-
The AI Supply Chain Is the New Attack Surface: From Ultralytics to LiteLLM
The Ultralytics YOLO supply chain attack in December 2024 compromised 60M+ downloads via GitHub Actions script injection, deploying cryptominers by injecting malicious code after code review but before…
-
Video.js v10: How One Developer Rewrote 16 Years of Code to Be 88% Smaller
Video.js v10 beta shipped March 2026 as a ground-up rewrite combining Video.js, Plyr, Vidstack, Media Chrome, and Mux Player into one project. The result: 88% smaller bundles, first-class…
-
NVIDIA Nemotron 3 Super: The Open-Weight Model That Beats GPT-4 on Code
NVIDIA released Nemotron 3 Super on March 12, 2026: a 120B-parameter open-weight model with 12B active parameters, hybrid Mamba-Transformer MoE architecture, and 1M-token context window. It tops DeepResearch…
-
OpenAI’s Workforce Doubles to 8,000: When a Research Lab Becomes an Enterprise Software Company
OpenAI plans to nearly double its workforce from 4,500 to 8,000 by December 2026, hiring 12 people per day. The expansion focuses on enterprise sales, technical ambassadorship, and…
-
Claudini: When AI Discovers Its Own Best Attacks
Claudini is an autonomous research pipeline where Claude Opus 4.6 iteratively designs adversarial attack algorithms against AI safety systems. It outperformed all 30+ human-designed methods, achieving 100% attack…
-
The Real Cost of Running AI in 2026: Compute, Revenue, and Who Can Actually Afford It
API costs dropped 93% but enterprise AI budgets are rising. Inference now accounts for 85% of AI spend, driven by agentic loops consuming 15x more tokens than chat.…
-
GitHub Will Train AI on Your Copilot Data Starting April 24. The Fine Print Is Worse Than the Headline.
GitHub announced on March 25, 2026 that starting April 24, all Copilot Free, Pro, and Pro+ interaction data will train AI models by default. The private repository loophole…
-
How Google TurboQuant Compresses LLM Memory by 6x With Zero Accuracy Loss
Google Research published TurboQuant on March 25, 2026: a KV cache compression algorithm that reduces LLM inference memory by 6x at 3-bit precision with zero accuracy loss and…
-
GPT-5.4 Pro Solved a Math Problem No Human Could Since 2019. Then a Supply Chain Attack Hit the AI Stack.
GPT-5.4 Pro solved an open math problem unsolved since 2019, verified independently by Epoch AI. FrontierMath scores jumped from 5% (GPT-4, 2024) to 50% (GPT-5.4 Pro, March 2026).…
-
Why OpenAI Killed Sora: $15 Million Per Day, a Dead Disney Deal, and the End of AI Video as a Consumer Product
OpenAI shut down Sora on March 24, 2026 after burning $15 million per day in inference costs against $2.1 million total lifetime revenue. Disney ended its $1 billion…
You must be logged in to post a comment.