Security
OpenAI's Agent Link Safety Isolates the Fetch, Not Prompt Injection
OpenAI's link-safety control stops quiet URL-based exfiltration by agents, not prompt injection. The trust boundary is moving from model output to network policy.
The Groundy archive · Page 26 of 43
All articles
Browse Groundy's complete archive of 1027 articles on AI, developer tools and infrastructure. Page 26 of 43.
The archive
601–624 of 1027 articles · Newest first
- Agents & Frameworks
Can LLM Agents Learn Cooperation Laws From Embodied Play?
LLawCo turns embodied agents' cooperation failures into readable rules fine-tuned into reasoning, making coordination policy an inspectable artifact engineers can edit.
- Security
No Verified 'React2Shell' Bulletin Exists: What Next.js Teams Should Check
A 'React2Shell' Vercel security bulletin is circulating, but no primary advisory, CVE, or technical write-up could be located as of 2026-06-29. Here is how to verify.
- Models & Research
Can Deep Learning Design RF Power Amplifiers Without Full EM Simulation?
A Chalmers/Tampere paper trains a CNN on EM-simulated layouts to search Doherty amplifier combiners in milliseconds. EM simulation is amortized, not eliminated.
- Security
Vercel on the Axios npm Compromise: Platform Scanning Has a Blind Spot
Vercel's Axios changelog exposes where platform defenses stop: post-publication egress blocks leave the install-time window on dev laptops and CI runners uncovered.
- Agents & Frameworks
Govern the Repo, Not the Agent: A New Risk Metric for AI-Native Code
A study of 930,000 agent-authored pull requests finds AI-native risk accumulates at the repository, not the agent, pushing governance to CI/CD and platform teams.
- Culture & Society
LLM-Generated VeriFast Specs Shift the Trust Bottleneck from Proofs to Review
An arXiv preprint tests LLM-generated VeriFast specs. The real danger is a silently accepted wrong contract, because verifiers treat any accepted spec as gospel.
Infrastructure & RuntimeGLM-5.2 on vLLM and Ascend: Open Weights Beyond NVIDIA
GLM-5.2 ships MIT-licensed with same-week serving recipes for NVIDIA vLLM and Huawei Ascend NPUs, breaking the open-weights-but-NVIDIA-only trade-off for self-hosters.
- Open Source
Hugging Face Is Absorbing Computer Vision Into Vision-Language Models
Computer vision is consolidating onto vision-language models on Hugging Face's Hub, so practitioners must prove each checkpoint does what its Model Card claims.
- Agents & Frameworks
Can an AI Agent Catch Cryptographic Misuse Before It Ships? Chai Tests the Claim
The Chai preprint reframes cryptographic misuse as a protocol-context recognition problem, claiming an LLM agent found a critical SSL-library flaw and 100-plus crypto bugs.
- Developer Tools
Vercel's CLI Is a Deployment Path, Not a Control Plane
Vercel's CLI is a deployment path, not a complete control plane. The April 2026 env-var breach and the push to agent operators make its lifecycle gaps impossible to ignore.
- Infrastructure & Runtime
How Vercel Runs Its Own CDN in Front of Discourse: A Self-Dogfooding Case Study
Vercel fronts its own Discourse forum with its CDN, but the edge cache only serves anonymous reads. Logged-in pages fall to the origin by design.
- Industry & Business
ByteDance's Doubao Seed 2.1 Pro: Production-Grade Claims, Vendor-Graded Evidence
ByteDance pitches Doubao Seed 2.1 Pro as production-grade AI at 6 CNY per million tokens, but scores are vendor-graded and Doubao is absent from independent leaderboards.
- Developer Tools
GLM-5.2 Goes Open Weights: What the Long-Horizon Coding Pitch Leaves Out
GLM-5.2 ships under MIT with real Coding Plan pricing, but the 1M context is opt-in, its coding benchmarks are vendor-reported, and Cursor integration is undocumented.
- Ethics, Policy & Safety
Medical AI Liability Needs a Clinical Harness
A June 2026 preprint reframes medical AI governance around runtime-governed clinical skills, exposing who is liable when diagnosis, scheduling, and documentation chain fails.
- Models & Research
Synthetic Clinical Notes from LLMs: Believable Prose Is Not Clinical Validity
A 70-patient synthetic EHR pipeline shows believable prose is not clinical validity. Fabricated dosages and labs poison models and invite HIPAA and FDA scrutiny.
Models & ResearchDoubao vs Qwen 3.7 vs GLM-5.2: Route by Axis, Not Leaderboard
Doubao-Seed-2.1 Pro's parity claims are vendor-graded and unreproduced. Route the Chinese flagship tier by axis: Qwen for ZH to EN, Doubao for EN to ZH, DeepSeek for code.
- Infrastructure & Runtime
Vercel Runtime Logs Surface CDN Cache Hits, Not the Eviction Cause
Vercel's Runtime Logs now show CDN cache key, tags, and revalidation reason per request, but the eviction cause stays hidden, forcing manual header inspection.
Models & ResearchCan Dynamic Experts Fix Catastrophic Forgetting in Robot Manipulation?
LiMoDE freezes a pre-trained expert bank and adds low-rank experts per new task, forcing continual robot learning to choose between architecture and replay.
- Developer Tools
HuggingFace Personal Copilot: The Bottleneck Is Your Codebase, Not Compute
Personal Copilot fine-tunes StarCoder on your file contents, not your commit history. The bottleneck is whether your code is clean enough to teach what Copilot does not know.
- Developer Tools
Llama 4 on Vercel's AI Model Gateway: Hosted Inference vs Self-Hosted vLLM
Vercel's AI Model Gateway promises zero-ops Llama 4 inference for Next.js apps but lists no models or rates. Self-hosting the 17B-active MoE keeps the knobs Vercel hides.
- Developer Tools
Vercel's Pre-Generate SSL Flow Stages Certs Before DNS Cutover
Vercel stages a Let's Encrypt cert via DNS-01 TXT records while traffic serves elsewhere, so HTTPS is valid at DNS cutover, though wildcards need Vercel nameservers.
- Models & Research
Error-Conditioned Neural Solvers vs Iterative Refinement: When Does Learned Correction Win?
A June 2026 preprint feeds the PDE residual into a neural corrector as input, not an optimization target, shifting surrogate cost from inference loops to training capacity.
- Models & Research
Vision-Language Models Move Past Object Detection: The MLLM Perception Shift
Vision-language models now reason over tables, charts, and documents, but detection-era benchmarks still rank them on box localization and undercount comprehension.
