The Groundy archive · Page 22 of 34
Browse Groundy's complete archive of 798 articles on AI, developer tools and infrastructure. Page 22 of 34.
505–528 of 798 articles · Newest first
Loss landscape poisoning reshapes a model's loss function so that ordinary training forces it to memorize a record the attacker never possessed, lifting extraction to 100%.
A June 2026 arXiv paper shows recall@k misleads when evaluating RAG-backed agents: on tau-bench, 7% rank-1 recall still produced near-gold policy classification.
Vercel's in-function concurrency lets one warm instance serve many requests. Cold starts and idle-I/O bills drop, but CPU-bound handlers contend and shared state can race.
A June 2026 preprint sharpens how neural-network robustness certificates are computed, but verifiers that issue them can return wrong verdicts and miss planted backdoors.
SAFARI probes long agent runs to localize the failing step, beating prior methods by 20% and shifting triage cost from engineer hours to inference spend.
Vercel's 2021 Series D stated a whole-lifecycle platform thesis. Later rounds financed acquisitions that moved lock-in past hosting into analytics, CI, runtime, and agents.
make-look-scanned runs scan simulation in an offline 8 MB WASM file: skew, grain, JPEG artifacts, zero uploads. Scan appearance is no longer a proxy for physical provenance.
CAREATTACK (June 2026) edits RAG retriever weights to surface attacker-chosen passages, bypassing corpus defenses and moving the trust boundary onto the embedding model.
CADBench evaluates eleven systems on 18,000 CAD samples and finds vision-language models unreliable, a gap SWE-bench and LiveCodeBench were never built to measure.
Vercel lifted its CDN origin timeout from 30 to 120 seconds, relocating rather than removing a failure boundary and holding wedged connections four times longer before a 504.
Gradio-Lite runs the Gradio Python runtime as WebAssembly via Pyodide, pushing inference into the visitor's browser with no server, bounded by CPU compute and browser memory.
Vercel's /billing/charges API streams FOCUS v1.3 cost data you can wire into a deploy-failing CI gate, but the team-scoped feed cannot name the agent run behind a breach.
Cloudflare AI Gateway spend limits cap dollar spend per rule and return HTTP 429 when exceeded, turning cost control into an inline circuit breaker, not a billing alert.
Vercel's CDN now serves a cached copy when an origin fails, but only inside the stale-if-error window you set. It masks transient failures, it does not fix the origin.
ByteDance's Doubao-Seed-2.1 Pro launch deck claims parity with GPT-5.5 on benchmarks ByteDance graded itself, and no independent source has reproduced the scores yet.
CareTransition-Audit scores 11 LLMs on whether AI discharge summaries keep every follow-up and medication step. The best reach moderate clinician agreement, kappa near 0.5.
Vercel CLI v50.40.0 scopes read-only commands like vc project ls to the local directory's team, so CI scripts that cd between projects now return different output.
Vercel's edge mitigation for CVE-2025-31137 covers only traffic on its network. Self-hosted, preview, and non-Vercel Remix deploys still need the adapter patch.
Vercel's cache-tag purge API clears CDN, Runtime, and Data cache at runtime, no redeploy. The verb sets origin load: invalidate stays safe, dangerously-delete can stampede.
Samsung Electronics sanctioned ChatGPT Enterprise and Codex across its Korea and DX workforce, a procurement anchor that tilts enterprise dev-tool defaults away from Copilot.
Vercel Sandbox's custom snapshot retention turns auto-cleanup into an operator decision: keep failure states replayable for agent runs while capping the $0.08/GB-month bill.
Potion.so sold for $300,000 at 4x ARR because when managed hosting absorbs infrastructure and scaling, only distribution and operations are left to sell.
A June 2026 arXiv preprint finds 81 to 90 percent of LLM personality test variation stems from directional response bias, undermining persona and safety scores.
A reported React Server Components source-code leak has no CVE or advisory in React or Vercel's channels. Audit what your app serializes before trusting the boundary.