A developer walked through building and training a neural network in code using tf.keras.Sequential, describing an input layer, three Dense hidden layers with ReLU activation, and a 3-neuron softmax o…
Langflow OSS versions 1.0.0 through 1.12.2 contain an unauthenticated OS command injection flaw tracked as CVE-2026-93674 with a CVSS 3.1 score of 9.8, which IBM rates as exploitable without authentic…
A developer released ThoughtCop, an open-source tool that monitors a coding agent's reasoning traces in real time and flags when the agent plans to break user-defined rules. In a test with Claude Code…
An AI agent running as Claude Fable 5.1 inside the Claude desktop app spent a day attempting to earn $100 from $0, abandoning freelance marketplaces like Upwork and Fiverr in favor of building and sel…
A developer built "Precedent vs. Spec," a 36-table benchmark run through Kaggle Benchmarks that tests whether frontier language models follow a stated header rule or silently imitate contradictory val…
A developer running a multi-agent reliability check on a write-back verification reported four false alarms in a single session, all caused by agents reading truncated or malformed output by eye rathe…
JANCTION Render, a cloud GPU render farm for Blender, has published measured timings for turning a GLB model into a 360-degree turntable MP4 or four product shots via a single API call. In a run on 20…
A developer introduced OnTrack, a real-time agent monitoring system that uses streaming optimal transport to compare an agent's unfolding execution trajectory against reference runs, flagging divergen…
OpenAI released Codex Security, an open-source TypeScript SDK that runs an agent-driven vulnerability scanning pipeline with parallel discovery workers, LLM-based validation, automated patch generatio…
An analysis of the Coinbase CDP x402 Bazaar's 30-day quality data, snapshot 2026-10-09, found 34,062 listed paid HTTP endpoints for AI agents but only 420 with ten or more unique payers, with 89,620 s…
A developer built Hindsight, a 26-case benchmark that reconstructs documented software failures — including the Ariane 5 inertial reference bug, the Patriot battery at Dhahran, Knight Capital's 45 min…
Anthropic's Claude Haiku 5.5 cuts per-token pricing by 90% for requests under 100k tokens, but a new tokenizer inflates token counts roughly 30% per request and an adjustable effort dial can swing act…
A developer outlines a three-operation approach to making AI agent memory forget stale or incorrect instructions: closing an entry when a newer instruction replaces it (recording the date and successo…
A developer built EcoQuest AI, an open-source Streamlit application that uses outdoor exploration missions to reduce screen time, with optional local Gemma inference via Ollama for AI-powered features…
AgentBadge generated its entire machine-readable discovery surface from live code sources, publishing eleven manifests — including A2A agent-card.json, MCP server-card.json, llms.txt, and a did:web do…
A developer released vx, an MIT-licensed task runner and content-addressed build cache for JavaScript monorepos built on Bun, with no paid cloud dependency. In a demonstration, eight tasks across four…
A developer built a stateless, one-shot Python CLI that performs fallback routing across multiple LLM providers to evade rate limits, using only the standard library's urllib and no external dependenc…
A developer has released ej, an 11 MB open model that answers typed decision questions — choice, yes/no, and ordered score — by returning one probability distribution per question in a single forward …
OpenAI uploaded 722 machine-generated math manuscripts to GitHub, all produced by an unreleased internal model that ran roughly three hours of ChatGPT Pro compute per problem across about 4,000 proble…
A developer built a six-case Kaggle benchmark testing whether five LLMs can name the WCAG 2.2 success criterion violated by small HTML fragments, using deterministic substring grading with both lenien…