René's URL Explorer Experiment


Title: Sid Ngeth's Blog · A blog about anything (but mostly development)

direct link

Domain: sngeth.github.io

Nonetext/html; charset=utf-8

Links:

Homehttps://sngeth.github.io/
Abouthttps://sngeth.github.io/about/
Podcastshttps://sngeth.github.io/podcasts/
Posts by Categorieshttps://sngeth.github.io/posts/
Reading Listhttps://sngeth.github.io/reading-list/
Readinghttps://sngeth.github.io/reading/
GitHub Profilehttps://github.com/sngeth
Sid Ngeth's Bloghttps://sngeth.github.io/
the second opinion needs to read the code Mar 29, 2026 llm agents code-review swe-bench research when GPT critiques Claude's patches without seeing the repo, the revised patches pass fewer tests. when GPT has the same... https://sngeth.github.io/llm/agents/code-review/swe-bench/research/2026/03/29/the-second-opinion-needs-to-read-the-code/
nanoclaw's memory architecture: sqlite + 200-message rolling window Mar 15, 2026 llm agents memory sqlite architecture a deep dive into nanoclaw's context retrieval mechanism -- no RAG, no embeddings, just SQLite with a cursor-based rolling window... https://sngeth.github.io/llm/agents/memory/sqlite/architecture/2026/03/15/nanoclaw-memory-architecture/
adding vim keybindings to opencode's tui Mar 11, 2026 opencode vim typescript solidjs tui a deep dive into adding vim keybindings to opencode's terminal UI -- matching claude code's vim feature set with a... https://sngeth.github.io/opencode/vim/typescript/solidjs/tui/2026/03/11/adding-vim-keybindings-to-opencode/
opencode permission rules: protecting your code from ai agents Mar 10, 2026 opencode ai git worktrees terraform a terraform destroy that wiped production https://sngeth.github.io/opencode/ai/git/worktrees/terraform/2026/03/10/opencode-permission-rules-protecting-git-worktrees/
server-side vs client-side tools in llm apis Jan 30, 2026 ai llm anthropic api-design when building agentic applications with the anthropic api, understanding the difference between client tools and server tools is essential. they... https://sngeth.github.io/ai/llm/anthropic/api-design/2026/01/30/server-vs-client-side-tools-in-llm-apis/
migrating to zig 0.15: the roadblocks nobody warned you about Oct 24, 2025 zig systems-programming breaking-changes i built a command-line security tool to analyze shell scripts before executing them (preventing those dangerous curl | bash situations).... https://sngeth.github.io/zig/systems-programming/breaking-changes/2025/10/24/zig-0-15-migration-roadblocks/
View All Posts →https://sngeth.github.io/posts
the second opinion needs to read the codehttps://sngeth.github.io/llm/agents/code-review/swe-bench/research/2026/03/29/the-second-opinion-needs-to-read-the-code/
the premisehttps://sngeth.github.io#the-premise
the experimenthttps://sngeth.github.io#the-experiment
run 1: the reviewer can’t read the codehttps://sngeth.github.io#run-1-the-reviewer-cant-read-the-code
run 2: the reviewer gets equal toolshttps://sngeth.github.io#run-2-the-reviewer-gets-equal-tools
what changedhttps://sngeth.github.io#what-changed
what the literature missedhttps://sngeth.github.io#what-the-literature-missed
what this meanshttps://sngeth.github.io#what-this-means
Claude Agent SDKhttps://docs.anthropic.com/en/docs/agents-and-tools/claude-code/sdk
OpenAI Codex CLIhttps://github.com/openai/codex
studies of LLM-generated code reviewshttps://arxiv.org/abs/2508.08661
Sharma et al. (2024)https://arxiv.org/abs/2310.13548
Codex CLIhttps://github.com/openai/codex
Panickssery et al., 2024https://arxiv.org/abs/2410.21819
Bavaresco et al., 2025https://arxiv.org/abs/2508.06709
Rethinking Mixture-of-Agentshttps://arxiv.org/abs/2502.00674
industrial-scale code review researchhttps://arxiv.org/abs/2505.17928
analysis of iterative AI code generationhttps://arxiv.org/abs/2506.11022
evaluations of Copilot’s security reviewhttps://arxiv.org/abs/2509.13650
switched to an agentic architecturehttps://github.blog/changelog/2026-03-05-copilot-code-review-now-runs-on-an-agentic-architecture/
produced enough false positiveshttps://github.blog/changelog/2025-10-28-new-public-preview-features-in-copilot-code-review-ai-reviews-that-see-the-full-picture/
spins up sandboxed environments per PRhttps://cloud.google.com/blog/products/ai-machine-learning/how-coderabbit-built-its-ai-code-review-agent-with-google-cloud-run
OpenAI Codex CLIhttps://github.com/openai/codex
SWE-bench Litehttps://www.swebench.com/lite.html
https://sngeth.github.io/llm/agents/code-review/swe-bench/research/2026/03/29/the-second-opinion-needs-to-read-the-code/index.html#disqus_thread
nanoclaw's memory architecture: sqlite + 200-message rolling windowhttps://sngeth.github.io/llm/agents/memory/sqlite/architecture/2026/03/15/nanoclaw-memory-architecture/
what is nanoclawhttps://sngeth.github.io#what-is-nanoclaw
the memory questionhttps://sngeth.github.io#the-memory-question
architecture overviewhttps://sngeth.github.io#architecture-overview
the 200-message rolling windowhttps://sngeth.github.io#the-200-message-rolling-window
cursor-based retrieval mechanismhttps://sngeth.github.io#cursor-based-retrieval-mechanism
accessing context beyond the windowhttps://sngeth.github.io#accessing-context-beyond-the-window
comparison to RAG-based systemshttps://sngeth.github.io#comparison-to-rag-based-systems
trade-offs and design decisionshttps://sngeth.github.io#trade-offs-and-design-decisions
source citationshttps://sngeth.github.io#source-citations
[email protected]https://sngeth.github.io/cdn-cgi/l/email-protection
https://sngeth.github.io/llm/agents/memory/sqlite/architecture/2026/03/15/nanoclaw-memory-architecture/index.html#disqus_thread
adding vim keybindings to opencode's tuihttps://sngeth.github.io/opencode/vim/typescript/solidjs/tui/2026/03/11/adding-vim-keybindings-to-opencode/
the requesthttps://sngeth.github.io#the-request
what claude code’s vim mode actually supportshttps://sngeth.github.io#what-claude-codes-vim-mode-actually-supports
opencode’s TUI architecturehttps://sngeth.github.io#opencodes-tui-architecture
the four-layer architecturehttps://sngeth.github.io#the-four-layer-architecture
cross-referencing against neovim’s sourcehttps://sngeth.github.io#cross-referencing-against-neovims-source
full vim mode referencehttps://sngeth.github.io#full-vim-mode-reference
how to try ithttps://sngeth.github.io#how-to-try-it
files changedhttps://sngeth.github.io#files-changed
more gotchas (found after shipping)https://sngeth.github.io#more-gotchas-found-after-shipping
escape key routing during generationhttps://sngeth.github.io#escape-key-routing-during-generation
#1764https://github.com/sst/opencode/issues/1764
#11111https://github.com/sst/opencode/issues/11111
sngeth/opencode (feat/vim-keybindings)https://github.com/sngeth/opencode/tree/feat/vim-keybindings
https://sngeth.github.io/opencode/vim/typescript/solidjs/tui/2026/03/11/adding-vim-keybindings-to-opencode/index.html#disqus_thread
opencode permission rules: protecting your code from ai agentshttps://sngeth.github.io/opencode/ai/git/worktrees/terraform/2026/03/10/opencode-permission-rules-protecting-git-worktrees/
let his claude code agent run terraform destroyhttps://alexeyondata.substack.com/p/how-i-dropped-our-production-database
https://sngeth.github.io/opencode/ai/git/worktrees/terraform/2026/03/10/opencode-permission-rules-protecting-git-worktrees/index.html#disqus_thread
server-side vs client-side tools in llm apishttps://sngeth.github.io/ai/llm/anthropic/api-design/2026/01/30/server-vs-client-side-tools-in-llm-apis/
anthropic documentationhttps://docs.anthropic.com/en/docs/build-with-claude/tool-use
web search documentationhttps://docs.anthropic.com/en/docs/build-with-claude/tool-use/web-search-tool
tool use with claudehttps://docs.anthropic.com/en/docs/build-with-claude/tool-use
web search toolhttps://docs.anthropic.com/en/docs/build-with-claude/tool-use/web-search-tool
https://sngeth.github.io/ai/llm/anthropic/api-design/2026/01/30/server-vs-client-side-tools-in-llm-apis/index.html#disqus_thread
Olderhttps://sngeth.github.io/page2

Viewport: width=device-width, initial-scale=1.0, maximum-scale=1

URLs of crawlers that visited me.