Are machines truly intelligent? AI researchers Subutai Ahmad and Nicolò Fusi join Doug Burger to compare transformer-based AI with the human brain, exploring continual learning, efficiency, and whether today’s models are on a path toward human intelligence.

Magic is an AI video editor that lets you place products into real filmed footage from locations like Paris, Times Square, and Tokyo. With 350+ templates, a hybrid real-footage+AI approach, and 90% first-try success, it's used by brands in 50+ countries. Pricing starts at $1 per video.

This issue covers Google's traumatized LLMs and DPO fix, DeepMind's cognitive taxonomy, UK's scaling law for AI cyberattacks, China's MERLIN for electronic warfare, and a sci-fi story.

DLSS 5 looks like a real-time generative AI filter for video games; OpenAI reportedly pivoting to a focus on business and productivity only; and more.

The article argues that AI progress, while significant, is better described as 'lossy self-improvement' rather than recursive self-improvement. Frictions such as narrow automatable research, diminishing returns from parallel agents, and resource bottlenecks suggest a more linear trajectory rather than exponential takeoff.

BankStatementLab launches on Product Hunt, offering AI-powered extraction of transactions from any bank statement PDF, converting them to Excel, CSV, or JSON. It works with any bank, format, or language, targeting accountants, bookkeepers, financial analysts, and businesses. Free tier includes 5 pages.

100+ AI models, one interface. You set the rhythm.

This article uses a coffee shop analogy to explain why AI companies like OpenAI and Anthropic are losing money despite high demand. It argues that as long as gross margins are positive, scaling up leads to profitability, contrasting Amazon's success with MoviePass's failure.

A comprehensive guide to securing Weaviate enterprise deployments using OIDC, RBAC, multi-tenant isolation, audit logging, and network security, illustrated through the fictional MedVector Health case study.

Despite incremental benchmark gains, GPT 5.4 in Codex offers real improvements in usability, speed, and context management, though Claude still wins on charm.

Anthropic's annualized revenue doubled in just two months, from $9B to $19B, surpassing expectations and indicating robust AI demand.

This week covers PostTrainBench showing AI agents can fine-tune LLMs but still lag humans; COVENANT-72B's distributed training achieving LLaMA2-level performance; Lean FRO's call for verification as AI writes more code; and CHMv2 highlighting computer vision challenges.

A summary and discussion of last week’s big AI news, including Perplexity's Personal Computer, Anthropic's code review tool, ChatGPT interactive visuals, Nvidia's Natron Free Super model, H200 halt for China, xAI cofounder departures, Anthropic lawsuit against Pentagon, and more.

Anthropic sues Trump administration over AI dispute with Pentagon; Musk's xAI restarts again; AI-generated fakes about war with Iran cause chaos online.

This article presents SPEX and ProxySPEX, algorithms that efficiently identify critical interactions in large language models from three perspectives: feature attribution, data attribution, and mechanistic interpretability. Leveraging structural properties like sparsity, low-degreeness, and hierarchy, these methods discover influential interactions between features, training data, and internal components with fewer ablations, demonstrating strong performance across long contexts, datasets, and model components.

This episode covers OpenAI's release of GPT-5.4 Pro and GPT-5.3 Instant, Google's Gemini 3.1 Flash Lite, Luma's unified multimodal models, and the escalating controversy over Anthropic being labeled a defense supply chain risk.

Turning fragmented knowledge into one simple system.

Microsoft Research open-sources AgentRx, a framework for automated diagnosis of AI agent failures. It pinpoints the first critical failure step using constraint synthesis and guarded evaluation, improving localization by 23.6% over baselines. The accompanying benchmark includes 115 annotated failed trajectories across three domains.

The article discusses the exponential growth of AI capabilities and its profound implications for work, markets, and policy. It describes the shift from 'co-intelligence' to 'managing AIs,' with AI agents like Claude Code and Codex now capable of completing complex tasks independently. The author demonstrates rapid AI progress through the 'Otter Test' and various benchmarks, noting that despite impressive capabilities, practical adoption is still early. The piece introduces StrongDM's radical 'Software Factory' experiment and the 'rolling disruption' caused by AI, including market volatility, corporate layoffs, and policy conflicts. Finally, the author warns about recursive self-improvement (RSI) potentially accelerating change, but emphasizes that the current window to shape AI's future…

This issue covers surprising AI progress acceleration, 14 metrics for AI R&D automation, an edge-computing traffic surveillance prototype in Bengaluru, a tiny satellite AI model for sea ice monitoring, ByteDance's CUDA-generating agent, and a sci-fi story about a drone war.

The U.S. Department of Defense designates Anthropic as a supply chain risk, while OpenAI's military deal sparks a 'cancel ChatGPT' movement. Also, ChatGPT 5.4 is hinted.

This is Part 2 of a 5-part series on fine-tuning sparse embeddings for e-commerce search. It covers training a SPLADE model on Modal's serverless GPUs using the Amazon ESCI dataset, including data loading, product text formatting, Modal setup, model creation, training function, SpladeLoss, YAML configuration, parallel hyperparameter sweeps, and a pitfall to avoid.

This article is Part 1 of a series on fine-tuning sparse embeddings for e-commerce search. It explains why dense embeddings fail in product search due to blurred exact matches, and how sparse embeddings preserve critical details. It introduces the SPLADE model, query expansion, and Qdrant's native sparse vector support. The fine-tuned system achieves a 29% improvement over BM25 on the Amazon ESCI dataset.

Discussion | Link

The 235th episode of Last Week in AI covers Anthropic's Sonnet 4.6 with 1M context and strong ARC-AGI-2 results, Google's Gemini 3.1 Pro, xAI's Grok 4.2 beta, and various tool updates. Business news includes Meta's up-to-$100B AMD chip deal, MatX raising $500M, World Labs $1B, and Simile $100M. Research highlights deep-thinking tokens, masking updates, and LLM attractor states. Policy discussions focus on Anthropic's stance on Pentagon contracts and distillation attacks.

A roundup of everything we shipped in February: Directory Snapshots for Sandboxes, a free GLM-5 endpoint, new billing API, and more.

OpenAI struck a deal with the Pentagon promising not to use its AI for fully autonomous weapons or mass surveillance of Americans, but critics argue the language is vague and may leave loopholes. Anthropic faced threats from the Trump administration for refusing similar terms. The article examines historical precedents and argues that Congressional legislation is the ultimate solution.

This issue covers an MIT-led paper on the economics of AGI, predicting humans will shift to verification; a study on LLMs boosting novice performance on bioweapon tasks; the GAMESTORE benchmark showing AIs underperform humans in video games; Physical Intelligence's robot deployments; and the Agents of Chaos study revealing AI agent fragility.

HAQQ Legal AI is now available on mobile, providing structured, jurisdiction-aware legal analysis with risk flags and exportable outputs, going beyond generic chatbot responses.

Learn how we built a production-ready, end-to-end RAG application in just 36 hours using the Query Agent and the new Weaviate Agent Skills library. The post explains the architecture, comparison with naive RAG, and step-by-step instructions.