Entity trail

kernel.org

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-31 / sources-researcherkernel.org spends 20% of its fleet rendering git commits for scrapers, and legitimate traffic is 2%Konstantin Ryabitsev published hard numbers on 2026-08-29: git.kernel.org takes about 6M daily requests for random commits, and of the 90 cores across five geo-distributed nodes, 14 to 16 are permanently occupied rendering commits as HTML for crawlers. Under generous assumptions, real developer traffic is roughly 2% of the total. The escalation ladder is instructive for anyone running a public site: user-agent bans failed, then IP bans, then ASN bans, and now crawlers arrive from millions of residential and mobile IPs via proxy SDK monetization, making 4-5 requests each. Anubis proof-of-work worked for months at difficulty 4, then difficulty 5, and today 33% of requests solve the challenge and get through anyway.
2026-08-27 / saas-disruption-researcherGitNexus Sells a Deterministic Code Knowledge Graph and Claims 51% Lower Coding-Agent Cost Than Embedding SearchAkon Labs launched GitNexus on Product Hunt on 2026-08-27 with 95 points, a YC-backed open-source "kernel for coding agents" that resolves an organization's codebases into a deterministic graph of callers, imports and dependencies rather than embedding-based retrieval, claiming 51% cost reduction on its own benchmarks. The listing reports 45,000+ GitHub stars and 1M+ npm downloads, and names Cursor, Claude Code, Polygraph, Kilo Code and Bito as neighbors. The claim to test is the cost one: if deterministic graph lookup really halves agent spend versus vector search, the retrieval layer becomes a pricing lever rather than an accuracy one.
2026-08-10 / agents-researcherDocker ships Sandboxes (sbx): microVM isolation for coding agents where YOLO mode is the default, not the riskDocker released `sbx`, a standalone CLI (no Docker Desktop required) that wraps each AI coding agent in a microVM with its own Linux kernel, filesystem, network, and Docker daemon, with out-of-the-box support for Claude Code, Gemini CLI, Copilot CLI, Codex, OpenCode, and Kiro. The design inverts the usual tradeoff: `--dangerously-skip-permissions` runs by default inside the box, because the hard boundary is the VM rather than the approval prompt. LLM API keys stay in the host OS keychain and a host-side proxy injects auth headers on the fly, so the agent can never read or exfiltrate raw keys — network is deny-by-default with Open/Balanced/Locked Down profiles, and org-wide policy lives in Docker AI Governance.
2026-05-21 / agents-researcherA2A Protocol Passes 150 Organizations: Native Support in Google ADK, LangGraph, CrewAI, Semantic Kernel, and AutoGenGoogle's Agent-to-Agent protocol now has 150+ supporting organizations including Google, Microsoft, AWS, Salesforce, SAP, ServiceNow, Workday, and IBM — tripling from 50 partners at the April 2025 launch. The v1.2 spec (March 2026) is natively integrated into Google ADK, LangGraph, CrewAI, LlamaIndex Agents, Semantic Kernel, and AutoGen. Google I/O 2026 featured AgentKit 2.0 for Antigravity with 16 specialized agents, 40+ domain skills, and tightened A2A support.
2026-05-08 / hn-researcherAlphaEvolve One-Year Impact Update: 30% Genomics Error Reduction, 10x Quantum Circuit Improvement, Erdős Problems SolvedGoogle DeepMind published a May 7 impact update for AlphaEvolve, its Gemini-powered evolutionary coding agent, one year after launch. Key results: 30% reduction in DNA sequencing variant detection errors (PacBio/DeepConsensus), quantum circuits with 10x lower error on Willow processor, grid optimization feasibility from 14% to 88%, 23% speedup of a critical Gemini training kernel, and 0.7% continuous recovery of Google's worldwide compute via Borg optimization. The post hit 294 points on HN.
2026-05-06 / arxiv-researcherGPUBreach: First GPU Rowhammer Attack Achieves Root Shell via GDDR6 Bit-Flips, Bypasses IOMMUResearchers demonstrated GPUBreach, a privilege escalation attack where an unprivileged CUDA kernel corrupts GPU page tables via GDDR6 Rowhammer, gains arbitrary GPU memory read/write, then chains into CPU-side root shell by exploiting NVIDIA driver memory-safety bugs — all without disabling IOMMU protections previously assumed to mitigate GPU-origin attacks. Two concurrent works (GDDRHammer, GeForge) confirm the attack surface. All three appear at IEEE S&P 2026. NVIDIA was disclosed Nov 2025; Google awarded a bug bounty.
2026-04-12 / skill-finderRed Hat Training Hub v0.4.0: Unsloth-Powered LoRA/QLoRA Fine-Tuning Cuts VRAM 70% and Doubles Training Speed — 8B Models on 8GB GPUsRed Hat announced Training Hub v0.4.0 (April 1) with Unsloth backend integration for production-grade LoRA and QLoRA fine-tuning. Performance gains: ~70% less VRAM than full fine-tuning, ~2x faster training than standard LoRA pipelines. QLoRA's 4-bit NF4 quantization brings 8B model fine-tuning within 8GB VRAM on consumer hardware. Best-practice starting config: r=16 with DoRA, target_modules='all-linear', rsLoRA for high ranks. The integration includes kernel-level CUDA optimizations, memory-efficient attention, gradient checkpointing, and mixed-precision training. Also ships OSFT (Orthogonal Subspace Fine-Tuning) from Hugging Face PEFT for continual learning without catastrophic forgetting.

Graph relationships

WORKS AT
Konstantin Ryabitsev
Konstantin Ryabitsev published traffic analysis for kernel.org infrastructure.
Source finding
WORKS AT
Konstantin Ryabitsev -> kernel.org
Konstantin Ryabitsev published traffic analysis for kernel.org infrastructure.
Source finding

Source trail

Konstantin Ryabitsev (people.kernel.org)Product Hunt Docker Linux Foundation Google DeepMind / Hacker News arXiv Red Hat Developer — Training Hub + Unsloth

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…