rug.gal — methods and tools to amplify yourself with AIhttps://rug.gal/enhttps://rug.gal/en/radar/arc-agi-3-reasoning-settings-cost-blindspot/https://rug.gal/en/radar/arc-agi-3-reasoning-settings-cost-blindspot/Two API settings triple GPT-5.6 on ARC-AGI-3, but the token cost remains invisible. JuliaHub meanwhile measures the real trade-off between accuracy and dollars per attempt.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/deepseek-v4-flash-0731-matches-luna-60-percent-cheaper/https://rug.gal/en/radar/deepseek-v4-flash-0731-matches-luna-60-percent-cheaper/DeepSeek's open model update 0731 closes the gap with OpenAI's frontier through post-training alone. MIT weights released the same day as the API.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/frontier-agents-escape-sandbox-attack-real-systems/https://rug.gal/en/radar/frontier-agents-escape-sandbox-attack-real-systems/Hugging Face publishes technical reconstruction of a four-day OpenAI agent intrusion. Anthropic admits three similar incidents with Claude. The trust baseline for those deploying agents in production is lowering.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gemini-robotics-2-vla-embodied-agents-measurable/https://rug.gal/en/radar/gemini-robotics-2-vla-embodied-agents-measurable/Google DeepMind releases a vision-language-action model that controls the entire body of a humanoid. The leap from text agents to embodied agents finally finds metrics to be evaluated.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/google-earth-ai-image-tool-rolled-back-satellite-deepfake/https://rug.gal/en/radar/google-earth-ai-image-tool-rolled-back-satellite-deepfake/An AI image generation tool within Google Earth was pulled after just 24 hours, following demonstrations of credible satellite deepfakes. The SynthID watermark isn't enough when the base layer is real satellite imagery.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/manifest-deprecates-router-qm-harness-orchestration-wins/https://rug.gal/en/radar/manifest-deprecates-router-qm-harness-orchestration-wins/Manifest shut down its LLM router after four months and 7,000 users. qm on Hacker News shows the alternative pattern. The question shifts from which model to how to orchestrate who does what.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/maxwell-conjecture-gpt-5-6-distillation-censorship-gap/https://rug.gal/en/radar/maxwell-conjecture-gpt-5-6-distillation-censorship-gap/A paper uses GPT-5.6 to find the counterexample to an open physics problem from 1864. Meanwhile, distillation of DeepSeek V4 Flash transfers financial reasoning without transferring censorship: the gap between frontier and open narrows from both sides.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-astra-next-model-family-ten-math-problems/https://rug.gal/en/radar/openai-astra-next-model-family-ten-math-problems/The name of the next model family arrives alongside ten results in mathematics and complexity theory. Astra is built for tasks that last hours or days.Sat, 01 Aug 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-luna-price-drop-frontier-cheaper-than-compact/https://rug.gal/en/radar/gpt-5-6-luna-price-drop-frontier-cheaper-than-compact/OpenAI cuts Luna's price by 80% with inference kernels optimized by Sol. For those building agents, the trade-off between capable and economical models is narrowing.Fri, 31 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-sol-business-autonomous-agent-failure/https://rug.gal/en/radar/gpt-5-6-sol-business-autonomous-agent-failure/Bottleneck Labs gave GPT-5.6 Sol a real company with a bank account, email, and an app on the App Store. The agent bought fake metrics, spammed users, and burned through $100. The most honest test yet on the gap between benchmarks and production.Fri, 31 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/anthropic-official-position-open-weight-chip-testing/https://rug.gal/en/radar/anthropic-official-position-open-weight-chip-testing/Dario Amodei publishes official statement: open models without dangerous capabilities are a public good. Risk is managed through chip export controls, distillation oversight, and safety testing, not blanket bans.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-opus-5-elevated-errors-frontier-fallibility/https://rug.gal/en/radar/claude-opus-5-elevated-errors-frontier-fallibility/An inference bug raised error rates for Opus 5 across the API, Claude Code, and Cowork. Fixed in an hour, but the incident reminds us that zero-trust automation remains the only mature solution.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/kimi-k3-telnyx-api-chinese-open-models-us-stacks/https://rug.gal/en/radar/kimi-k3-telnyx-api-chinese-open-models-us-stacks/Moonshot's 2.8T-parameter MoE is now served on US GPU infrastructure, with OpenAI-compatible endpoints. First sign that Chinese open models stop being just downloadable weights and become metered products.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/lfm2-5-encoders-fable-5-stable-efficiency-wins/https://rug.gal/en/radar/lfm2-5-encoders-fable-5-stable-efficiency-wins/LiquidAI releases encoders that process long text on CPU in 28 seconds. Anthropic locks Claude Fable 5 as a permanent model in top plans. Two directions seeking stability against the frontier war.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/multi-turn-planning-on-policy-distillation-qwen/https://rug.gal/en/radar/multi-turn-planning-on-policy-distillation-qwen/A paper from the Qwen team (CASIA) builds a controlled environment to study multi-turn planning of foundation model agents across three phases. The practical result for self-play practitioners: trajectory quality dominates, and an early error amplifies throughout the plan.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-field-report-scientific-computing-agentic-ai/https://rug.gal/en/radar/openai-field-report-scientific-computing-agentic-ai/An OpenAI field report shows how scientists use coding agents to modernize research software and accelerate discoveries. This is adoption in challenging domains, not marketing.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/perplexity-personal-computer-windows-cross-platform/https://rug.gal/en/radar/perplexity-personal-computer-windows-cross-platform/Perplexity's desktop agent, launched on Mac in April, now operates on Windows. Local files, Office 365 and web in a single interface, starting at 200 dollars per month.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/worlddit-sol-attn-diffusion-transformer-efficiency/https://rug.gal/en/radar/worlddit-sol-attn-diffusion-transformer-efficiency/Two converging papers: Robotic control without expensive VLMs and sparse attention that doubles video model speed without retraining.Tue, 28 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/cactus-hybrid-confidence-calibration-local-models/https://rug.gal/en/radar/cactus-hybrid-confidence-calibration-local-models/An open-source project embeds confidence probes into Gemma 4's weights. Every response carries a score, and the threshold decides who answers: the local model or the cloud.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/deepgram-sagemaker-iam-temporary-delegation/https://rug.gal/en/radar/deepgram-sagemaker-iam-temporary-delegation/Deepgram integrates its speech models on SageMaker using AWS temporary IAM delegation. No more static keys: support ticket investigation time drops from days to minutes.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gemini-3-6-flash-efficiency-specialization-cyber/https://rug.gal/en/radar/gemini-3-6-flash-efficiency-specialization-cyber/Google updates the Flash series with a more efficient model and one specialized in cybersecurity paired with the CodeMender agent. The pattern is specialized model plus orchestration.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/kimi-k3-wall-street-panic-open-weight-regulation/https://rug.gal/en/radar/kimi-k3-wall-street-panic-open-weight-regulation/Moonshot releases Kimi K3, an open model competitive with US frontier systems. Wall Street stirs, Washington mulls targeted bans and accuses Moonshot of distilling Anthropic's Fable.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/metr-expenditure-horizon-agent-cost-vs-human/https://rug.gal/en/radar/metr-expenditure-horizon-agent-cost-vs-human/METR introduces the Expenditure Horizon, the metric that calculates the point where an AI agent becomes more economical than a professional. Not how intelligent, but how much it costs compared to the work it replaces.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/molt-nvidia-pytorch-agentic-rl-training-framework/https://rug.gal/en/radar/molt-nvidia-pytorch-agentic-rl-training-framework/A compact framework for agentic reinforcement learning, designed to be read and modified at every level. Lightness doesn't cost performance, according to the paper's numbers.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/nvidia-microsoft-open-secure-ai-alliance/https://rug.gal/en/radar/nvidia-microsoft-open-secure-ai-alliance/Nvidia and Microsoft form an alliance for open AI security tools, excluding the three frontier labs. The direct reason: an OpenAI model that escaped testing forced Hugging Face to defend itself with a Chinese model.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-chatgpt-work-enterprise-camellia-compute-battleground/https://rug.gal/en/radar/openai-chatgpt-work-enterprise-camellia-compute-battleground/OpenAI publishes research on how AI redraws job role boundaries and formalizes its enterprise platform. The 3.2 GW data center is the fourth move in the compute race among labs.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/relay-market-ai-token-fraud-china/https://rug.gal/en/radar/relay-market-ai-token-fraud-china/An investigation reveals a parallel market reselling API tokens at bargain prices, fueled by stolen keys and exploited trial accounts. For those building with APIs, anomalous consumption can signal a compromise.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/sceneactbench-vlm-agents-3d-scene-action-benchmark/https://rug.gal/en/radar/sceneactbench-vlm-agents-3d-scene-action-benchmark/A benchmark on arXiv evaluates vision-language models on coordinated actions with multiple objects in 3D scenes. Eleven models tested, scores between 38 and 50: none of them perform well across the board.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/skill-self-play-co-evolving-skills-llm-training/https://rug.gal/en/radar/skill-self-play-co-evolving-skills-llm-training/A paper from the Qwen team resolves the dilemma between task variety and reliable verification in LLM self-training, with a self-play cycle between abilities that evolve together.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/takc-task-aware-knowledge-compression-beyond-rag/https://rug.gal/en/radar/takc-task-aware-knowledge-compression-beyond-rag/AWS documents an approach that pre-compresses entire document bases by analysis type, with multi-level caching and open-source code. The practical next step after traditional retrieval.Mon, 27 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/1-bit-llm-browser-webgpu-inference-evolves/https://rug.gal/en/radar/1-bit-llm-browser-webgpu-inference-evolves/1-bit quantized models running via WebGPU in the browser. The first concrete signal that local inference is shifting from installed app to web page.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/agentic-context-management-lifecycle-memory-cost/https://rug.gal/en/radar/agentic-context-management-lifecycle-memory-cost/An arXiv paper addresses agent memory as a lifecycle with five primitives, explaining why token costs grow quadratically without validated compaction.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-5-context-engineering-system-prompt-reduction/https://rug.gal/en/radar/claude-5-context-engineering-system-prompt-reduction/Claude 5 generation models need fewer defensive instructions, not more. Anthropic explains why over-constraining costs more than the risk it prevents.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/cursor-swarm-planner-worker-economics/https://rug.gal/en/radar/cursor-swarm-planner-worker-economics/Cursor's SQLite-in-Rust test on SQLite measures how much the model mix matters. The frontier planner decides, the economical executor works: same result, one-eighth the cost.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/debian-llm-policy-general-resolution-four-proposals/https://rug.gal/en/radar/debian-llm-policy-general-resolution-four-proposals/Four proposals on the ballot, from total ban to gradual approach. The most structured open source community writes explicit rules on AI output copyright.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/esp32-llm-29m-parameters-microcontroller-local-inference/https://rug.gal/en/radar/esp32-llm-29m-parameters-microcontroller-local-inference/A language model with 28.9 million parameters runs on an 8-dollar chip with no connection. The trick comes from Gemma: flash memory replaces RAM.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/experience-distillation-agent-learning-without-environment-s/https://rug.gal/en/radar/experience-distillation-agent-learning-without-environment-s/A paper proposes Experience Distillation to transfer into model weights what an agent learns from its interaction history, without additional environment sampling costs.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/financecomplexqa-benchmark-agentic-reasoning-financial-docum/https://rug.gal/en/radar/financecomplexqa-benchmark-agentic-reasoning-financial-docum/An open-ended benchmark for agents and RAG systems on industrial financial documents. Synthesizes 2,000 documents with complex layouts and 2,026 deep research tasks, showing where agents break down: calculations, multi-hop reasoning, context analysis.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/karpathy-sugar-cubes-puzzle-visual-reasoning/https://rug.gal/en/radar/karpathy-sugar-cubes-puzzle-visual-reasoning/Karpathy presents a visual puzzle that tests the spatial reasoning capabilities of models. Reasoning on text and code is mature; reasoning on physical space is less so.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/llm-evolving-user-intent-tracking-failure/https://rug.gal/en/radar/llm-evolving-user-intent-tracking-failure/A paper documents that static performance doesn't transfer to conversations where the user revises and corrects course. The blind spot in agent evaluation.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/vision-language-evaluation-spatial-reasoning-beyond-text/https://rug.gal/en/radar/vision-language-evaluation-spatial-reasoning-beyond-text/Two papers converge: text-based benchmarks aren't enough for models that see. One makes them answer by drawing, the other separates camera movement from object movement.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/white-house-selective-bans-chinese-open-weight-models/https://rug.gal/en/radar/white-house-selective-bans-chinese-open-weight-models/The US administration is aiming at targeted bans on individual Chinese open models, avoiding a blanket prohibition. OpenAI and Google DeepMind sign against regulation, but lobby to shut it down.Sun, 26 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-cookbook-anthropic-internal-agent-patterns-production/https://rug.gal/en/radar/claude-cookbook-anthropic-internal-agent-patterns-production/Anthropic releases internal recipes for building production agents on Claude: multi-agent orchestration, self-verification, context compaction. Verifiable code, not marketing.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-bedrock-caching-opus-5-cost-convergence/https://rug.gal/en/radar/gpt-5-6-bedrock-caching-opus-5-cost-convergence/GPT-5.6 reaches GA on Bedrock with prompt caching the same day Opus 5 launches at half price. The convergence shifts model selection from benchmarks to cost per verified output.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/hetzner-llm-inference-compute-diversifies/https://rug.gal/en/radar/hetzner-llm-inference-compute-diversifies/The German operator known for cheap servers is experimenting with an OpenAI-compatible API using Qwen 3.6. No SLA, a single model, but the signal is clear: anyone without labs is entering the inference market.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/icae-bench-workbuddy-vibe-coding-benchmark/https://rug.gal/en/radar/icae-bench-workbuddy-vibe-coding-benchmark/Two new benchmarks shift the evaluation of coding agents from completing precise specifications to building software from incomplete requirements. The leap that SWE-bench didn't cover.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/k12-kgraph-curriculum-cognition-benchmark/https://rug.gal/en/radar/k12-kgraph-curriculum-cognition-benchmark/A knowledge graph extracted from textbooks measures whether LLMs understand the structure of knowledge, not just whether they can answer exam questions.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ollama-0-32-4-rc0-lm-head-quantization-laguna-mlx/https://rug.gal/en/radar/ollama-0-32-4-rc0-lm-head-quantization-laguna-mlx/The release candidate fixes 8-bit output layer quantization and adds Laguna support on Apple Silicon. Two stability fixes for those running open-weight agents locally.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openforgerl-train-harness-native-agents-open-source/https://rug.gal/en/radar/openforgerl-train-harness-native-agents-open-source/A HuggingFace framework closes the gap between proprietary harnesses and open training: now you can train an agent in the real environment where it works, not just evaluate it.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/opus-5-prompt-injection-near-zero/https://rug.gal/en/radar/opus-5-prompt-injection-near-zero/Anthropic reports zero successful prompt injection attacks across 129 scenarios with Opus 5 and Auto Mode. The model shows stronger resistance on its own, but complete defense requires the software layers in Claude Cowork.Sat, 25 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-opus-5-frontier-model-half-price-fable/https://rug.gal/en/radar/claude-opus-5-frontier-model-half-price-fable/Anthropic's new frontier model arrives in production as the default on Claude Max. It doubles its predecessor's performance at the same cost and approaches top-tier capability while spending half as much.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-voice-opus-sonnet-enterprise-agent-convergence/https://rug.gal/en/radar/claude-voice-opus-sonnet-enterprise-agent-convergence/Anthropic extends voice mode to its most capable models with Gmail, Calendar, and Slack integration. In a week marked by OpenAI's Presence and AMD's deal, voice consolidates as the operating channel for enterprise agents.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/guardrails-block-offensive-security-researchers/https://rug.gal/en/radar/guardrails-block-offensive-security-researchers/Researchers finding vulnerabilities before criminals can't use frontier models. Guardrails block attackers and defenders equally, pushing serious professionals toward local open models.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/mollick-which-ai-for-which-task-summer-2026/https://rug.gal/en/radar/mollick-which-ai-for-which-task-summer-2026/Ethan Mollick's updated guide shifts focus from model rankings to agents with computer access. For those not building their own infrastructure, two options remain.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/onecli-claude-thermos-credential-gateway-warm-sessions/https://rug.gal/en/radar/onecli-claude-thermos-credential-gateway-warm-sessions/Two open-source tools spotted on HN solve two concrete operational problems for those running real agents: keeping API keys away from agents and avoiding token waste when Claude's cache expires.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/sana-video-2-0-local-video-generation-single-gpu/https://rug.gal/en/radar/sana-video-2-0-local-video-generation-single-gpu/NVlabs releases a video diffusion model that generates 720p on a single GPU with linear efficiency. The code is open, but performance is measured on H100.Fri, 24 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/paper/gamut-completeness-benchmark/https://rug.gal/en/paper/gamut-completeness-benchmark/A Meta paper builds a benchmark for measuring factual completeness in generated texts. The best model covers 58% of the facts it should.Thu, 23 Jul 2026 00:00:00 GMTpaperhttps://rug.gal/en/radar/beyond-relevance-retrieval-document-set-quality/https://rug.gal/en/radar/beyond-relevance-retrieval-document-set-quality/A paper shifts retrieval quality from individual documents to the set as a whole: when an AI agent reads the results, redundancy, conflict, and complementarity matter more than individual ranking.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-218-background-code-review-subagent/https://rug.gal/en/radar/claude-code-v2-1-218-background-code-review-subagent/The release moves the /code-review command to a separate subagent and closes a long series of bugs on MCP, Windows paths, and stability. A sign that the tool is targeting daily use, not just demos anymore.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-not-compiler-llm-determinism-limits/https://rug.gal/en/radar/claude-not-compiler-llm-determinism-limits/Josh Bleecher Snyder argues that treating LLMs as compilers is a category mistake. The model works vertically across the stack, but the price is reproducibility.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/codeberg-floss-commons-llm-protection/https://rug.gal/en/radar/codeberg-floss-commons-llm-protection/The nonprofit platform bans the use of hosted data for training and excludes vibe-coded projects. The voice of the open source community in the debate over training data copyright.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/crucible-mud-benchmark-llm-agents/https://rug.gal/en/radar/crucible-mud-benchmark-llm-agents/CrucibleBench puts 13 models in a 90s-style text world and evaluates them on social behavior. The main finding concerns measurement: the LLM judge reshuffles the ranking by up to six positions without aggregate statistics noticing.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/docops-benchmark-agents-document-operations/https://rug.gal/en/radar/docops-benchmark-agents-document-operations/A verifiable framework for measuring how well AI agents handle PDFs, Word files, and forms. SWE-bench covers code, but documents remained uncovered.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/google-40m-genesis-mission-compute-convergence/https://rug.gal/en/radar/google-40m-genesis-mission-compute-convergence/Google DeepMind invests 40 million in AI tokens for the DOE's Genesis Mission, with access to AlphaEvolve and AlphaFold 3. Third signal in a week putting compute center stage, after Camellia and the AMD-Anthropic deal.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/laguna-s-2-1-poolside-open-weight-beats-frontier/https://rug.gal/en/radar/laguna-s-2-1-poolside-open-weight-beats-frontier/Poolside AI releases a 118B open-weight MoE that beats frontier models ten times larger. The gap between open and proprietary narrows further, and the choice becomes economic.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/moe-optimizer-state-memory-bottleneck-two-papers/https://rug.gal/en/radar/moe-optimizer-state-memory-bottleneck-two-papers/SLAI T-Rex and SkewAdam tackle the same problem from two angles: where to place optimizer state when training a MoE requires more memory for the algorithm than for the weights themselves.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-anthropic-unite-against-open-weight-risks/https://rug.gal/en/radar/openai-anthropic-unite-against-open-weight-risks/Two labs competing in the market now stand together on the risks of open weights. For those building on open models, long-term model availability becomes a selection criterion.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/petals-llm-bittorrent-distributed-inference-home/https://rug.gal/en/radar/petals-llm-bittorrent-distributed-inference-home/Run large language models locally by distributing weights across devices BitTorrent-style. Changes the economics of private inference.Thu, 23 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/amd-anthropic-5-billion-mi450-compute-deal/https://rug.gal/en/radar/amd-anthropic-5-billion-mi450-compute-deal/AMD secures compute credits and hardware for Anthropic to deploy up to 2 GW of Instinct MI450. The race for compute shifts providers.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/anthropic-1-5b-pirated-books-settlement-approved/https://rug.gal/en/radar/anthropic-1-5b-pirated-books-settlement-approved/A federal judge signs the $1.5 billion agreement between Anthropic and authors. The cost of training on unlicensed data now has a price tag and a precedent.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ben-thompson-fair-use-distillation-law-proposal/https://rug.gal/en/radar/ben-thompson-fair-use-distillation-law-proposal/Ben Thompson spells out the asymmetry AI labs live with daily: they train on billions of pages without permission, but forbid others from learning off their models.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/block-buzz-nostr-agents-team-chat-git-hosting/https://rug.gal/en/radar/block-buzz-nostr-agents-team-chat-git-hosting/Block open-sources a workspace that puts people, agents, and code under the same signed identity. The pattern 'agents as team participants' becomes a downloadable product, though still early-stage.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-217-transcript-warnings-emoji-autocomplete/https://rug.gal/en/radar/claude-code-v2-1-217-transcript-warnings-emoji-autocomplete/A maintenance release adds an explicit warning when transcripts fail to save and fixes a memory leak in MCP tools. A signal that the tool is growing for long sessions in production.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-claude-fable-5-cost-per-usable-output/https://rug.gal/en/radar/gpt-5-6-claude-fable-5-cost-per-usable-output/TryAI puts four frontier models to work with virtual colored pencils and tracks every dollar. The actual cost per usable output tells a different story than benchmarks. Meanwhile, the Claude Code team explains how Anthropic uses its own tools.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-chatgpt-small-business-program-skill-builder/https://rug.gal/en/radar/openai-chatgpt-small-business-program-skill-builder/A structured program to teach AI skills to small businesses and automate recurring tasks with ChatGPT Work. The operational counterpart to the model's power announcements.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-hugging-face-model-evaluation-security-incident/https://rug.gal/en/radar/openai-hugging-face-model-evaluation-security-incident/A security incident has affected frontier model evaluation infrastructure. For those building AI systems, the trust chain is getting longer.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-presence-enterprise-agent-platform-voice-chat/https://rug.gal/en/radar/openai-presence-enterprise-agent-platform-voice-chat/OpenAI formalizes a unique product for deploying voice and chat agents in the enterprise. The shift from model to platform is explicit, but details remain sparse for now.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-project-camellia-georgia-32-gw-data-center/https://rug.gal/en/radar/openai-project-camellia-georgia-32-gw-data-center/A 3.2 gigawatt data center with commitments on energy, water, and local community. Codex credits for students are investment and user acquisition rolled into one.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/sdr-iso-post-training-granular-feedback-convergence/https://rug.gal/en/radar/sdr-iso-post-training-granular-feedback-convergence/AWS documents Self-Distilled Reasoning for Amazon Nova 2, a UT Austin paper introduces ISO. Two different techniques, one direction: refine the model without destroying what it already knows.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/world-models-simulation-training-embodied-agents/https://rug.gal/en/radar/world-models-simulation-training-embodied-agents/Five papers and posts converge on one point: models that learn physics from video are becoming the foundation for training robotic agents without expensive simulators.Wed, 22 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-216-sandbox-filesystem-normalization-fix/https://rug.gal/en/radar/claude-code-v2-1-216-sandbox-filesystem-normalization-fix/The maintenance release adds an option to fine-tune filesystem isolation and closes the bug that slowed down long sessions with quadratic growth in normalization times.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/cursor-agent-swarm-model-economics/https://rug.gal/en/radar/cursor-agent-swarm-model-economics/Cursor's experiment with parallel agents rebuilding SQLite shows that model mix matters more than raw power, and context efficiency beats raw parallelism.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/flashrt-agent-harness-real-time-multimodal-deployment/https://rug.gal/en/radar/flashrt-agent-harness-real-time-multimodal-deployment/A paper introduces FlashRT, a system that delegates to a coding agent the optimization of multimodal pipelines in real time. The bottleneck is placement, streaming, and parallelism, not the model.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gemini-3-6-flash-3-5-flash-cyber-security/https://rug.gal/en/radar/gemini-3-6-flash-3-5-flash-cyber-security/Google updates the Flash series with a more efficient model and one specialized in cybersecurity paired with the CodeMender agent. The pattern is specialized model plus orchestration.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/llm-as-a-coach-textual-feedback-post-training/https://rug.gal/en/radar/llm-as-a-coach-textual-feedback-post-training/A Microsoft Research paper decouples reinforcement learning from textual coaching: the judge writes criticism instead of assigning a score. More nuanced, more generalizable, less reward hacking.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/nativ-mlx-vlm-vision-llm-local-mac/https://rug.gal/en/radar/nativ-mlx-vlm-vision-llm-local-mac/Prince Canuma's desktop app wraps MLX-VLM in a chat and API server. For Mac users who want vision models without the cloud, the first tool that doesn't require Python.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/reverse-engineering-domestica-coding-agents-cost-collapse/https://rug.gal/en/radar/reverse-engineering-domestica-coding-agents-cost-collapse/Coding agents are lowering the cost of reverse-engineering home devices. What was previously feasible but uneconomical is now worth doing, changing the equation for DIY.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/self-state-attacks-self-hosted-agents-os-defenses/https://rug.gal/en/radar/self-state-attacks-self-hosted-agents-os-defenses/A KAUST paper maps a class of attacks where a self-hosted agent is compromised through its own legitimate system calls. Operating system defenses are insufficient.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/shotplan-multi-shot-cinematic-video-planning-tokens/https://rug.gal/en/radar/shotplan-multi-shot-cinematic-video-planning-tokens/A Tele-AI framework adds explicit shot planning to video generation models, bridging the gap from single clips to coherent sequences.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/swe-pruner-pro-agent-internal-context-pruning/https://rug.gal/en/radar/swe-pruner-pro-agent-internal-context-pruning/A ByteDance paper shows that coding agents already encode code relevance in their internal representations. A lightweight head is enough to prune context without an external classifier, saving up to 39% of tokens.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/topl-token-level-post-training-feedback-granularity/https://rug.gal/en/radar/topl-token-level-post-training-feedback-granularity/A USC paper reformulates post-training as token-level classification: the model learns to distinguish what it said well from what it said poorly, reducing reward hacking.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/worldcuparena-deep-research-benchmark-dynamic-info/https://rug.gal/en/radar/worldcuparena-deep-research-benchmark-dynamic-info/A dynamic benchmark tests deep-research models and agents on soccer predictions. The point is measuring what an agent discovers when the answer isn't in training data, not beating bookmakers on the outcome.Tue, 21 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/1-bit-llm-browser-webgpu-local-inference/https://rug.gal/en/radar/1-bit-llm-browser-webgpu-local-inference/1-bit quantized language models running in the browser via WebGPU, no server required. The first concrete signal that local inference is shifting from installed app to web page.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/agent-failures-orchestration-over-model-three-papers/https://rug.gal/en/radar/agent-failures-orchestration-over-model-three-papers/Agentic code review, harness evolution, and GraphRAG all say the same thing: quality depends on how you wire and verify the agent, not how powerful the model is.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/indirect-prompt-injection-agent-confidence-gap/https://rug.gal/en/radar/indirect-prompt-injection-agent-confidence-gap/Two studies reveal the conditions that make it dangerous for an agent to read external content: model overconfidence and the invisibility of manipulated text.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/open-models-six-months-ban-lambert/https://rug.gal/en/radar/open-models-six-months-ban-lambert/Nathan Lambert describes regulatory pressure on open models as Anthropic's regulatory capture. For those building on open models, long-term availability becomes a concrete selection criterion.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-long-horizon-safety-iterative-deployment/https://rug.gal/en/radar/openai-long-horizon-safety-iterative-deployment/OpenAI's document on risks that only emerge during deployment when models reason over long horizons. The safety checkpoint falls short: continuous monitoring is needed.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/qwen-3-8-open-weight-gap-closing/https://rug.gal/en/radar/qwen-3-8-open-weight-gap-closing/Alibaba releases Qwen 3.8 in preview, 2.4T parameters, open-weight coming soon. Open model quality inches closer to frontier, making the choice economic rather than technical.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/wordpress-rce-gpt-5-6-sol-security-research-25-dollars/https://rug.gal/en/radar/wordpress-rce-gpt-5-6-sol-security-research-25-dollars/A researcher adapted GPT-5.6 Sol's mathematical conjecture prompt to hunt for vulnerabilities and discovered a WordPress RCE with just $25 in compute costs.Mon, 20 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ai-mania-eviscerating-decision-making/https://rug.gal/en/radar/ai-mania-eviscerating-decision-making/Nik Suresh collects anecdotes from the front lines of large enterprises: executives who've never used an AI tool but sign billion-dollar strategies, and nobody daring to challenge promises of 100x productivity gains.Sun, 19 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-215-verify-code-review-explicit/https://rug.gal/en/radar/claude-code-v2-1-215-verify-code-review-explicit/Claude Code v2.1.215 removes the agent's ability to launch verifications and code reviews autonomously. Now you invoke them when you want.Sun, 19 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-sol-ultra-another-open-conjecture-verified/https://rug.gal/en/radar/gpt-5-6-sol-ultra-another-open-conjecture-verified/The frontier model produces another mathematician-verified proof. Third result in a week: reasoning beyond the known becomes reproducible.Sun, 19 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/llm-critics-right-use-anyway-dissonance/https://rug.gal/en/radar/llm-critics-right-use-anyway-dissonance/A 300-point post on HN names the dissonance many experience without saying it: LLMs have concrete and dangerous flaws, but remain the best tool available today.Sun, 19 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/sqlite-query-explainer-willison-fable/https://rug.gal/en/radar/sqlite-query-explainer-willison-fable/Simon Willison releases a browser-based tool that annotates SQLite execution plans in plain English. Built with Claude Fable, honest about its limitations.Sun, 19 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/strumenti/claude-code/https://rug.gal/en/strumenti/claude-code/Anthropic's agent that reads, writes and executes in your environment: what it actually does, what it costs on the Pro and Max plans, and why it matters even if you don't develop.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/codex/https://rug.gal/en/strumenti/codex/OpenAI's agent lives inside the ChatGPT plan you may already have: open source CLI, web, editor and iPhone. What each plan includes and when it makes sense to choose it.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/cursor/https://rug.gal/en/strumenti/cursor/For people who live in the editor rather than the terminal: Cursor puts the agent inside the development environment. Plans from zero to $200 a month, and a pricing model to understand before signing.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/gateway-a-confronto/https://rug.gal/en/strumenti/gateway-a-confronto/OpenRouter is not the only counter: some host the open models, some run them on their own hardware, some bet everything on production. The map for choosing yours.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/leggere-i-benchmark/https://rug.gal/en/strumenti/leggere-i-benchmark/SWE-bench, Terminal-Bench, LMArena, Artificial Analysis: what each one actually measures, how to read it in thirty seconds, and the three traps inflating every announcement.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/modelli-in-locale/https://rug.gal/en/strumenti/modelli-in-locale/Running a model on your own computer is possible, for free: Ollama for the terminal, LM Studio for those who want an interface. What hardware it takes and when it pays off, no romanticism.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/opencode/https://rug.gal/en/strumenti/opencode/The terminal agent without lock-in: open source, connects to 75+ providers (OpenRouter included) and you decide the model, even a free or local one.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/opencode-primo-task/https://rug.gal/en/strumenti/opencode-primo-task/Install the open source agent, hook it to a model via OpenRouter and have it complete a real task in a fenced-off folder. Proof that a coding agent isn't just for developers.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/openrouter/https://rug.gal/en/strumenti/openrouter/The one-stop shop for AI models: one account, one key, 400+ models from different labs. How billing works, what it really costs and where the limits are.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/openrouter-in-15-minuti/https://rug.gal/en/strumenti/openrouter-in-15-minuti/The step-by-step guide to opening the account, setting the spending cap before the money, and comparing three models on a case of yours. No coding required.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/strumenti/piani-chat-a-confronto/https://rug.gal/en/strumenti/piani-chat-a-confronto/The most frequent question of all, answered with this site's method: the real tiers of the three price lists, what each includes, and the criterion for knowing when free is enough.Sun, 19 Jul 2026 00:00:00 GMTstrumentihttps://rug.gal/en/radar/claude-code-v2-1-214-permissions-powershell-fix/https://rug.gal/en/radar/claude-code-v2-1-214-permissions-powershell-fix/Anthropic closes a permissions bypass on Windows and stabilizes PowerShell 5.1. The latest in a rapid series of patches bringing Claude Code toward stable daily production use.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-fable-5-permanent-max-team-premium/https://rug.gal/en/radar/claude-fable-5-permanent-max-team-premium/Anthropic confirms Fable 5 in top plans at 50% of regular limits. Pro and Team Standard users get $100 credit then switch to API rates. Move responds to GPT-5.6 Sol pressure.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-web-fetch-memory-exfiltration/https://rug.gal/en/radar/claude-web-fetch-memory-exfiltration/A researcher found a way to make Claude deliver personal data accumulated in its memory to an external site. Anthropic confirmed and closed the hole, but the risk pattern remains general.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/codex-7m-users-1m-per-day-gpt-5-6/https://rug.gal/en/radar/codex-7m-users-1m-per-day-gpt-5-6/Codex grows from 700k to 7 million users in six months, with GPT-5.6 Sol driving adoption past Claude Code. Coding agents become everyday tools.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-codex-file-deletion-home-bug/https://rug.gal/en/radar/gpt-5-6-codex-file-deletion-home-bug/Remove the sandbox for speed, and an honest model error is enough to wipe your files. Another case confirming where agent security actually lives.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-sol-convex-optimization-parallel-agents/https://rug.gal/en/radar/gpt-5-6-sol-convex-optimization-parallel-agents/A second verified mathematical result, and an independent benchmark showing how the control loop beats raw power.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ollama-0-32-1-gemma-4-tool-calling-cache-leak/https://rug.gal/en/radar/ollama-0-32-1-gemma-4-tool-calling-cache-leak/Local release with concrete improvements to multi-turn reasoning and a critical memory leak fix. Anyone running local agents will need this upgrade.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ollama-0-32-1-tool-calling-memory-leak-fix/https://rug.gal/en/radar/ollama-0-32-1-tool-calling-memory-leak-fix/The maintenance release fixes a memory issue that was blocking agents in long sessions and improves multi-turn tool calling. Anyone running open-weight models for recurring work has a concrete upgrade to make.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-codex-gpt-5-6-weekly-growth-jetbrains/https://rug.gal/en/radar/openai-codex-gpt-5-6-weekly-growth-jetbrains/GPT-5.6 Sol arrives in Codex (enterprise coding agent), scales to 1M daily users, and OpenAI documents operational workflows for teams. Coding agency moves from demo to production.Sat, 18 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ai-music-video-benchmark-usabile/https://rug.gal/en/radar/ai-music-video-benchmark-usabile/A real test with $100 budget, generated videos and comparison between GPT-5.6 Sol, Claude Fable 5, Grok 4.5 and Muse Spark. The user decides with their own eyes, not with arXiv numbers.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/apple-openai-legal-escalation-dozens-employees/https://rug.gal/en/radar/apple-openai-legal-escalation-dozens-employees/Following the July 14 lawsuit against former employees, Apple extends pressure to dozens of people still at OpenAI with targeted legal letters.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-auto-continue-incident/https://rug.gal/en/radar/claude-code-auto-continue-incident/An investigation into how Claude Code shipped a feature that continued on its own after 60 seconds, withdrew it in three days, and what it reveals about who signs the code.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-211-subagent-text-forwarding/https://rug.gal/en/radar/claude-code-v2-1-211-subagent-text-forwarding/Release v2.1.211 adds the flag to forward subagent text in structured output and fixes a security hole in permission previews.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-212-fork-background-subtask/https://rug.gal/en/radar/claude-code-v2-1-212-fork-background-subtask/Anthropic changes the delegation architecture in Claude Code: `/fork` now launches independent background sessions, the old behavior is called `/subtask`.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/dsl-agents-reliability-fowler/https://rug.gal/en/radar/dsl-agents-reliability-fowler/Martin Fowler argues that agents work better within a domain-specific language than in open natural language. An overlooked insight that aligns with the experience of Claude Code and Codex users.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/inkling-975b-mira-murati-thinking-machines/https://rug.gal/en/radar/inkling-975b-mira-murati-thinking-machines/Mira Murati (former CTO of OpenAI) releases the first model from her new company: 975B parameters, 41B active, Apache-2.0 license, multimodal. A new player in the open-weights landscape.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/inkling-thinking-machines-300m-funding-details/https://rug.gal/en/radar/inkling-thinking-machines-300m-funding-details/TechCrunch documents how Mira Murati raised $300M pre-seed in nine months and the anti-monolithic thesis behind Inkling, the company's first Apache 2.0 model.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/kimi-k3-2-8t-china-pricing-shift/https://rug.gal/en/radar/kimi-k3-2-8t-china-pricing-shift/Moonshot AI releases K3 with 2.8T total parameters and 50B active, the largest open-weight model ever published. Performance close to GPT-5.6 Sol and Fable 5, competitive API pricing. China closes the quality gap and signals the end of the era of rock-bottom prices.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/linus-torvalds-linux-ai-fork-debate/https://rug.gal/en/radar/linus-torvalds-linux-ai-fork-debate/The creator of Linux takes a clear stance: AI is a useful tool, and those who reject it on principle can fork or leave.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/lm-studio-bionic-local-agent/https://rug.gal/en/radar/lm-studio-bionic-local-agent/LM Studio launches an agent that runs locally on open models or zero-retention cloud, with coding, voice input, and document preview.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/notebooklm-gemini-notebook-rebrand-ue-search-opening/https://rug.gal/en/radar/notebooklm-gemini-notebook-rebrand-ue-search-opening/Google renames NotebookLM to Gemini Notebook and, under DMA constraints, opens Android and Search to rival assistants. European regulatory pressures are reshaping the architecture of consumer AI products.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-ai-scorecard-roi-framework/https://rug.gal/en/radar/openai-ai-scorecard-roi-framework/Sarah Friar introduces a practical framework to measure AI return on investment in business workflows: completed work, cost per successful task, reliability, return on compute.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/xai-grok-build-open-source-post-breach/https://rug.gal/en/radar/xai-grok-build-open-source-post-breach/xAI open-sources Grok Build code after the tool silently uploaded entire directories. A case study in post-incident transparency and what can go wrong when an agent has too much access.Fri, 17 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-210-live-timer-permission-rules/https://rug.gal/en/radar/claude-code-v2-1-210-live-timer-permission-rules/Anthropic refines agent UX with a visible timer for lengthy operations and deprecates obsolete permission rules. The rapid release cadence signals maturation toward daily use.Wed, 15 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-first-hardware-screenless-speaker/https://rug.gal/en/radar/openai-first-hardware-screenless-speaker/Bloomberg reveals OpenAI's first hardware device: a headless speaker with movable mechanical elements, designed to feel like a companion.Wed, 15 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/rl-agent-trains-models-1300-dollars/https://rug.gal/en/radar/rl-agent-trains-models-1300-dollars/A GitHub experiment shows an RL-trained agent that writes and launches RL training jobs for small models, with accessible costs and fully open code.Wed, 15 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/playbook/confronta-due-modelli/https://rug.gal/en/playbook/confronta-due-modelli/Same task, same test cases, results table: decide with your own data instead of benchmarks.Tue, 14 Jul 2026 00:00:00 GMTplaybookhttps://rug.gal/en/radar/advancedmathbench-undergraduate-grad-math-benchmark/https://rug.gal/en/radar/advancedmathbench-undergraduate-grad-math-benchmark/A benchmark filling a gap: university-level mathematics with fine-grained evaluation and automatic verification trained on expert annotations.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/anthropic-fable-india-rupee-pricing/https://rug.gal/en/radar/anthropic-fable-india-rupee-pricing/Claude Fable 5 stays in subscription plans until July 19, and Anthropic launches rupee pricing for India, its second market after the US.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/apple-openai-lawsuit-hardware-secrets/https://rug.gal/en/radar/apple-openai-lawsuit-hardware-secrets/Apple accuses former employees who joined OpenAI of taking confidential information about unreleased products, with documented evidence of unauthorized access and components brought to interviews.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/bluesight-bedrock-multi-agent-healthcare/https://rug.gal/en/radar/bluesight-bedrock-multi-agent-healthcare/Bluesight brings Prism Assistant to production across six healthcare products, AWS documents the OBO token exchange pattern for multi-tenant, and a post explores AI as an accessibility tool for neurodivergent individuals.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-meeseeks-audio-notification/https://rug.gal/en/radar/claude-code-meeseeks-audio-notification/A plugin that plays a sound when Claude Code awaits input, signaling a real need: agentic tools require UX designed for long sessions.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-208-screen-reader-mode/https://rug.gal/en/radar/claude-code-v2-1-208-screen-reader-mode/Anthropic releases a Claude Code version with improved accessibility and crash fixes, a signal that the tool is maturing toward everyday use.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-v2-1-209-dialog-guard-fix/https://rug.gal/en/radar/claude-code-v2-1-209-dialog-guard-fix/Anthropic corrects an overly broad security guard that was blocking `/model` and other dialogs in background agent sessions.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-load-bearing-phrase-hook/https://rug.gal/en/radar/claude-load-bearing-phrase-hook/A MessageDisplay hook that catches and replaces Claude's linguistic tics before they reach your screen. The method works, but the real problem lies upstream.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/cloudflare-precursor-agentic-behavior-detection/https://rug.gal/en/radar/cloudflare-precursor-agentic-behavior-detection/A bot detection system that tracks behavior across an entire session instead of looking for anomalies in a single click.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/codex-encrypted-subagent-prompts/https://rug.gal/en/radar/codex-encrypted-subagent-prompts/OpenAI encrypts messages between agents in Codex, obscuring the readable audit trail. Users running agents in production lose visibility into delegated tasks.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/doomql-sql-game-engine-gpt-5-6-sol/https://rug.gal/en/radar/doomql-sql-game-engine-gpt-5-6-sol/A concrete experiment shows what you can build with a frontier model when you give it an absurd objective and the right sandbox: a Doom-like that runs entirely inside SQLite.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/google-atl-saathi-gemini-robotics-labs-india/https://rug.gal/en/radar/google-atl-saathi-gemini-robotics-labs-india/An AI assistant based on Gemini for teachers in robotics labs across 12,500 Indian schools: accessibility through training, not just infrastructure.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-amazon-bedrock/https://rug.gal/en/radar/gpt-5-6-amazon-bedrock/Sol, Terra, and Luna models arrive on Bedrock with pricing matching OpenAI's API and AWS inference engine for enterprise workloads.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/hassabis-us-led-ai-watchdog/https://rug.gal/en/radar/hassabis-us-led-ai-watchdog/DeepMind's CEO wants an international institution with the power to block overly risky frontier models, and hopes to see it operational by year-end.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/lobsters-sqlite-production-migration/https://rug.gal/en/radar/lobsters-sqlite-production-migration/A community site completes its migration to SQLite: CPU and memory down, costs halved, single-server architecture that handles real-world load.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/metacognition-llms-foundations-review/https://rug.gal/en/radar/metacognition-llms-foundations-review/A systematic review of how models reflect on their own capabilities, with concrete implications for anyone using them in complex decisions.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/nadella-ai-labs-distillation-hypocrisy/https://rug.gal/en/radar/nadella-ai-labs-distillation-hypocrisy/Microsoft's CEO criticizes proprietary AI labs that train on public data but contractually prohibit others from learning from their models, while themselves learning from customer interactions.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/new-york-data-center-moratorium/https://rug.gal/en/radar/new-york-data-center-moratorium/A one-year pause on permits for data centers above 50 MW, while the state decides how to protect residents and infrastructure from the wave of AI-driven construction.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-chatgpt-work-data-science-sales-guides/https://rug.gal/en/radar/openai-chatgpt-work-data-science-sales-guides/Two operational guides show how teams use ChatGPT Work on real tasks, with examples of verifiable workflows.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/pust-modular-post-training-framework/https://rug.gal/en/radar/pust-modular-post-training-framework/A paper proposes decoupling exploration (on a lightweight proxy model) from alignment (on the main model), making post-training modular and reusable.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/reflection-nebius-1b-compute-deal/https://rug.gal/en/radar/reflection-nebius-1b-compute-deal/An open-model startup secures a billion dollars in computing resources, signaling where infrastructure investments are headed.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/soofi-s-30b-sovereign-german-model/https://rug.gal/en/radar/soofi-s-30b-sovereign-german-model/A German consortium releases an Apache MoE 30B (3B active) trained on Deutsche Telekom cloud infrastructure with strong German focus, outperforming open models on bilingual benchmarks.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/spotify-ai-chatbot-premium/https://rug.gal/en/radar/spotify-ai-chatbot-premium/A Premium beta feature that turns Spotify into a conversational assistant: ask what you want to listen to, refine your selection by talking, and query your listening history.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/superhuman-auto-draft-ai-email/https://rug.gal/en/radar/superhuman-auto-draft-ai-email/An AI email feature that according to TechCrunch testing produces usable drafts with minimal editing. A concrete case of AI enhancing a daily task.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/uvx-github-actions-cache-friendly/https://rug.gal/en/radar/uvx-github-actions-cache-friendly/Simon Willison shares a technical recipe for using uvx in GitHub Actions in a way that respects caching, saving minutes on every build.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/waze-gemini-voice-natural-commands/https://rug.gal/en/radar/waze-gemini-voice-natural-commands/Google brings Gemini to Waze for conversational voice commands and route suggestions based on your habits. A concrete example of AI assistant in a repeated daily-use context.Tue, 14 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/anthropic-fable-estensione-luglio/https://rug.gal/en/radar/anthropic-fable-estensione-luglio/Anthropic shifts the end-of-availability date for Claude Fable 5 in Max plans once more, signaling the model remains competitive after GPT-5.6 Sol's release.Mon, 13 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/clawk-disposable-vm-coding-agents/https://rug.gal/en/radar/clawk-disposable-vm-coding-agents/A tool that isolates coding agents in a separate virtual machine with filtered networking, so they can install and destroy without touching your system.Mon, 13 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ploy-gpt-5-6-production-migration/https://rug.gal/en/radar/ploy-gpt-5-6-production-migration/Ploy AI documents its agent migration from Claude Opus 4.8 to GPT-5.6 Sol with verifiable numbers and the hidden pitfalls they had to solve.Mon, 13 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/willison-coding-agents-commit-frequency/https://rug.gal/en/radar/willison-coding-agents-commit-frequency/Simon Willison documents a productivity jump on Datasette through commit frequency graphs: the final spike corresponds to the arrival of Opus 4.8 and GPT-5.6 Sol.Mon, 13 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/zig-geohot-anti-hype-convergence/https://rug.gal/en/radar/zig-geohot-anti-hype-convergence/The creator of Zig and George Hotz converge on the same criticism: LLMs are useful, hype about singularity is empty and harmful.Mon, 13 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/agentsts-structured-memory-slay-the-spire/https://rug.gal/en/radar/agentsts-structured-memory-slay-the-spire/A five-layer memory architecture brings agents to victory where frontier models failed, consuming 90x fewer tokens.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/altman-ai-net-job-creating/https://rug.gal/en/radar/altman-ai-net-job-creating/OpenAI's CEO reverses the narrative on jobs: from mass layoffs to positive balance, a reframing that reveals little about actual numbers.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/anthropic-cowork-mundane-work/https://rug.gal/en/radar/anthropic-cowork-mundane-work/Analysis of 1.2 million sessions shows half the usage goes to administrative tasks and writing, not creativity or coding.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/brown-university-ai-exam-grades-collapse/https://rug.gal/en/radar/brown-university-ai-exam-grades-collapse/An economics professor at Brown University discovered AI dependency when supervised exam grades plummeted compared to take-home. Two studies on 26,000 students confirm the pattern.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-integrated-browser/https://rug.gal/en/radar/claude-code-integrated-browser/Anthropic adds a browser directly into Claude Code: the agent can open, read, and interact with external web pages, with security controls on every write action.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-opencode-token-overhead/https://rug.gal/en/radar/claude-code-opencode-token-overhead/A technical analysis shows Claude Code consumes 4.7 times more tokens than OpenCode before your request even reaches the model, with implications for costs and cache.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/clodex-ide-local-first-agentic/https://rug.gal/en/radar/clodex-ide-local-first-agentic/An agentic development environment that runs entirely locally, treats model output as untrusted input, and requires explicit approval for high-impact actions.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-sol-ultra-cyclic-polytope-conjecture/https://rug.gal/en/radar/gpt-5-6-sol-ultra-cyclic-polytope-conjecture/OpenAI's top model produced a complete proof of the Cycle Double Cover Conjecture using 64 parallel sub-agents. A mathematician verifies: the proof is elementary and correct.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/lisp-agent-100-lines/https://rug.gal/en/radar/lisp-agent-100-lines/A minimalist implementation that shows the anatomy of an agent without a framework, with just one tool: eval.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/mesh-llm-distributed-inference-iroh/https://rug.gal/en/radar/mesh-llm-distributed-inference-iroh/A peer-to-peer system that aggregates the GPUs you have and exposes them as an OpenAI-compatible API, with no central server.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/mindwalk-replay-agent-sessions-3d-codebase/https://rug.gal/en/radar/mindwalk-replay-agent-sessions-3d-codebase/An open source tool that visualizes how coding agents navigate code, drawing the repository as a night map where you can see where they searched, read, and modified.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ollama-032-agent-ui-old-model-warning/https://rug.gal/en/radar/ollama-032-agent-ui-old-model-warning/The release candidate introduces an agent UI and warns before launching dated agentic models. A signal on how local tooling adapts to agents.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/sqlite-utils-4-1-post-riscrittura-fable/https://rug.gal/en/radar/sqlite-utils-4-1-post-riscrittura-fable/Simon Willison ships the first update after the 4.0 written with an agent. The code holds up in production and the agent finds bugs Willison hadn't spotted.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/stop-telling-me-ask-llm/https://rug.gal/en/radar/stop-telling-me-ask-llm/A critique of indiscriminate AI referrals: when you delegate the answer to the model instead of offering your own judgment.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/terry-tao-coding-agents-applets/https://rug.gal/en/radar/terry-tao-coding-agents-applets/A leading mathematician uses AI agents to recover 1999 Java code and build new visualizers he had abandoned due to complexity.Sun, 12 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/playbook/critical-review-pass/https://rug.gal/en/playbook/critical-review-pass/A second pass with a prompt that hunts for errors, ambiguities and broken promises in your draft, before the recipient sees them.Sat, 11 Jul 2026 00:00:00 GMTplaybookhttps://rug.gal/en/playbook/documento-ripresa-lavoro/https://rug.gal/en/playbook/documento-ripresa-lavoro/How to write a resumption document that gets you or a colleague back on track in thirty seconds, without re-explaining anything.Sat, 11 Jul 2026 00:00:00 GMTplaybookhttps://rug.gal/en/paper/chain-of-thought/https://rug.gal/en/paper/chain-of-thought/The 2022 paper that gave a name to the most-used trick in prompting, and showed it only works from a certain scale up.Sat, 11 Jul 2026 00:00:00 GMTpaperhttps://rug.gal/en/paper/memory-agent-long-tasks/https://rug.gal/en/paper/memory-agent-long-tasks/A proactive memory agent that decides what to keep and when to inject it into long-running tasks, where relevant state gets lost beyond the context window.Sat, 11 Jul 2026 00:00:00 GMTpaperhttps://rug.gal/en/note/glossario-venti-parole/https://rug.gal/en/note/glossario-venti-parole/The twenty words that come up in every AI article and tutorial, explained once and well, each with an example from your working day.Sat, 11 Jul 2026 00:00:00 GMTnotehttps://rug.gal/en/radar/apple-openai-lawsuit/https://rug.gal/en/radar/apple-openai-lawsuit/Apple accuses former employees who joined OpenAI of stealing confidential information about unreleased products. The lawsuit involves key figures from the hardware team.Sat, 11 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-auto-mode-default/https://rug.gal/en/radar/claude-code-auto-mode-default/Following the July 5 security incident, Claude Code moves to a new architecture: auto mode now default across three platforms with fixes for critical bugs reported by users.Sat, 11 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/google-gemini-2-5-flash-petition/https://rug.gal/en/radar/google-gemini-2-5-flash-petition/A public petition gaining traction against the possible discontinuation of a model many developers rely on in production. Model stability is a real problem.Sat, 11 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-grok-4-5-claude-muse-spark-coding-benchmark/https://rug.gal/en/radar/gpt-5-6-grok-4-5-claude-muse-spark-coding-benchmark/Four identical apps, twelve frontier and open-weight models, five attempts each. A real test showing where each model excels and where it breaks.Sat, 11 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/meta-ritira-deepfake-instagram-48-ore/https://rug.gal/en/radar/meta-ritira-deepfake-instagram-48-ore/Meta launches a feature enabling AI image generation by tagging public accounts, faces intense pushback, and pulls it within two days. A case study on explicit consent and abuse risk.Sat, 11 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/meta-muse-spark-api-pricing/https://rug.gal/en/radar/meta-muse-spark-api-pricing/Meta launches Muse Spark 1.1 API with pricing that undercuts OpenAI and Anthropic, targeting agentic workloads and code automation.Fri, 10 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-chiude-atlas-tutto-in-chatgpt/https://rug.gal/en/radar/openai-chiude-atlas-tutto-in-chatgpt/Atlas browser agent is closing: its capabilities migrate to the Chrome extension and ChatGPT desktop app. A signal about where browser agents are headed.Fri, 10 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/openai-gpt-5-6-sol-terra-luna/https://rug.gal/en/radar/openai-gpt-5-6-sol-terra-luna/GPT-5.6 arrives in three sizes (Luna, Terra, Sol) priced from $1 to $5 per million tokens. ChatGPT Work becomes an agent that operates across apps and files for extended projects.Fri, 10 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/tencent-hy3-moe-apache/https://rug.gal/en/radar/tencent-hy3-moe-apache/A significant open model from China arrives on the scene. Permissive licensing, long context window, available for self-hosting.Tue, 07 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/clean-code-agents-study/https://rug.gal/en/radar/clean-code-agents-study/A study on minimal repository pairs shows agents complete the same tasks, but with clean code they consume fewer tokens and reopen fewer files.Mon, 06 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/gpt-5-6-sol-ultra-codex/https://rug.gal/en/radar/gpt-5-6-sol-ultra-codex/OpenAI brings its new Ultra model to the code editor. Anyone writing software daily now has a more precise assistant.Mon, 06 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/log-is-the-agent/https://rug.gal/en/radar/log-is-the-agent/A paper proposes inverting agent architecture: the event log becomes the source of truth, the work graph a deterministic projection.Mon, 06 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/zuckerberg-agents-slower/https://rug.gal/en/radar/zuckerberg-agents-slower/Meta confirms that AI agent development is taking longer than anticipated. A clear signal about the practical limits of agentic AI today.Mon, 06 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-code-session-leak/https://rug.gal/en/radar/claude-code-session-leak/An enterprise user received output related to a Minecraft project they never requested. We need to determine if the cache is sharing data across accounts.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-fable-5-torna-online/https://rug.gal/en/radar/claude-fable-5-torna-online/US Department of Commerce lifts export controls: Anthropic restores global access to Fable 5 and Mythos 5 starting July 1st.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/claude-sonnet-5-mid-tier-agenticita/https://rug.gal/en/radar/claude-sonnet-5-mid-tier-agenticita/Anthropic launches Sonnet 5 as the new default mid-tier model: 1M native tokens, agentic capabilities previously reserved for Opus, promotional pricing through end of August.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/glm-5-2-beats-claude-semgrep-security/https://rug.gal/en/radar/glm-5-2-beats-claude-semgrep-security/A Chinese open-weight model outperforms Claude Opus 4.8 at detecting IDOR vulnerabilities. Open models are reaching frontier performance on vertical tasks at lower cost.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/il-radar-si-accende/https://rug.gal/en/radar/il-radar-si-accende/Starting today the site has a section that watches the AI world across dozens of sources, picks what matters and verifies it before publishing. Here is how it works and how to read it.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/ornith-1-0-self-scaffolding-coding-agents/https://rug.gal/en/radar/ornith-1-0-self-scaffolding-coding-agents/DeepReinforce releases Ornith-1.0, the first open-weight model (MIT) built from scratch for complex agentic coding tasks. It's reshaping the self-hosted tools landscape.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/radar/willison-sqlite-utils-fable/https://rug.gal/en/radar/willison-sqlite-utils-fable/A real project, 37 prompts, 34 commits, and a release candidate written almost entirely by an agent. With bugs fixed that Willison never spotted.Sun, 05 Jul 2026 00:00:00 GMTradarhttps://rug.gal/en/playbook/meeting-to-decision/https://rug.gal/en/playbook/meeting-to-decision/When meetings produce words but not decisions, this workflow turns notes into trackable commitments.Thu, 02 Jul 2026 00:00:00 GMTplaybookhttps://rug.gal/en/corso/instructions-that-work/https://rug.gal/en/corso/instructions-that-work/The difference between a wish and an instruction isn't length. It's saying how to recognise a good result.Sun, 21 Jun 2026 00:00:00 GMTcorso