Terminal-Bench-Science introduces a benchmark of 70 verifiable scientific workflows across five disciplines; the leading agent resolves only 30% of tasks. HN discusses whether its tests capture correctness and instruction-following, and the risks and promise of AI-assisted research.
A free, framework-free Colab curriculum teaches practical LLM engineering from raw APIs, covering RAG, agents, evals, fine-tuning, security, and serving. HN discussion focuses on whether the basics are useful and on the importance of evaluation harnesses.
Experiential is an open-source Rust gateway that unifies hosted, BYOK, and local models, then uses traces and simulated evaluations to optimize model routing for agent workflows. HN discussion focuses on caching, routing tradeoffs, fine-tuning, telemetry, and how it differs from LiteLLM and similar gateways.
An investigation finds at least 19 US law-enforcement agencies used more than $920,000 in opioid settlement funds for AI-powered Flock, Peregrine, and other surveillance systems. Critics say the spending diverts addiction-treatment money into controversial mass-policing infrastructure.
The article argues Silicon Valley is ignoring growing public opposition to generative AI, AI data centers, surveillance cameras, and AI glasses. HN debates whether the backlash reflects legitimate economic and privacy concerns, poor industry messaging, or resistance to technological change.
A benchmark compares small language models for pocket-scale, on-device inference. HN discusses the gap between benchmark scores and real-world usefulness, plus mobile RAM, power, and NPU constraints that still limit local LLM deployment.
The article compares US household spending on LLM subscriptions with World of Warcraft to question AI’s claimed mass adoption. HN debates conflicting subscriber estimates, enterprise usage, churn, and whether consumer subscriptions can sustain LLM costs.
Anthropic is previewing MHS, a model-agnostic standard for AI agents to discover and control lab and industrial hardware through shared drivers, safety metadata, and MCP-compatible interfaces. HN debates whether it meaningfully advances existing automation protocols and whether LLMs are safe abstractions for physical equipment.
Google launched Gemini 3.5 Transcribe for low-latency streaming and recorded-audio transcription, with formatting, custom vocabulary, multilingual support, and speaker attribution. HN users compared it with Whisper, Soniox, Voxtral, and ElevenLabs, finding strong accuracy but ongoing concerns around latency, diarization, pricing, and privacy.
An AI-assisted fuzzer found a division-by-zero crash in FFmpeg, though commenters note the issue may have been discovered previously by OSS-Fuzz. The discussion focuses on whether LLM-built fuzzing tools provide meaningful novelty beyond conventional fuzzers.
hrdx is a Go terminal multiplexer built around running multiple coding agents across persistent project workspaces. HN discussion compares its agent-focused workflow and API with tmux and similar tools.
A developer says sustained pressure to use Claude Code and multiple parallel agents has left them unable to understand or write code independently. HN discusses productivity pressure, skill atrophy, mental health, delegation, and quality risks.
Google’s Gemini Omni 1.1 Flash adds scene extension, keyframe interpolation, video references, faster low-resolution drafts, and 4K upscaling through its API. HN discusses whether better controllability will matter more than raw quality, alongside creative-industry disruption, deepfake concerns, and Google’s video strategy.
An open-source project ports Cursor’s Grok Bot desktop app to Linux by rebuilding its Electron runtime and native modules. It provides unofficial access to Grok bots and the cloud computer, though Cursor and xAI do not support the build.
Tare is a local Claude Code skill that analyzes request logs to explain quota usage, context overhead, projects, tools, and runaway automation. HN users compare it with built-in status lines and discuss how large contexts and subagents rapidly consume quotas.
A Lean repository formalizes the solution to the Hopf problem. HN discussion focuses on the growing role of LLMs in producing and checking large-scale mathematical proofs, and what human understanding means when formalization is machine-assisted.
AI agents can edit Excel files but often cannot reliably recalculate or validate complex workbooks without Excel, sometimes masking errors with Python estimates. HN discusses spreadsheet-agent limitations, benchmark fairness, and whether the promoted Python-based Orcaset alternative adds enough value.
Small, fast models are becoming capable enough for routine coding, tool use, and business workflows at a fraction of frontier-model costs. HN discusses the trade-offs among model quality, latency, harness design, local hosting, and whether cheaper inference could unlock new AI products.
Salem Robotics develops an AI-plus-classical-robotics software layer for hazardous industrial inspections, using existing robot hardware for tasks such as nuclear surveys and leak detection. The discussion explores the practical boundary between learned perception and explicit geometric planning in safety-critical robots.
Opslane is an open-source agent that watches web sessions and error data, prioritizes user-impacting bugs, investigates them, and opens verified pull requests. HN discussion highlights its agent workflow and substantial privacy concerns around session recording.