Ask HN: What are tools you have made for yourself since the advent of AI?
TL;DR Highlight
HN developers share personal tools built with AI: sandboxes, MCP, QA automation
Who Should Read
Developers using AI coding agents who want ideas for personal tools to build
Core Mechanics
Community Reactions
- Several agent sandbox tools were shared
- sandvault runs agents in a separate macOS user account.
- clodpod uses a macOS VM and claude-sandbox uses podman.
- yoloai launches agents with max permissions in docker/podman etc. and extracts changes via git diffs.
- Let the LLM fill only a JSON contract, one commenter said
- An amp-sim preset is filled by the LLM as small JSON, then a script generates the XML for the plugin.
- The same contract/adapter pattern is used for Terraform, game engine presets and CI pipelines.
- Agents got confused between docs and code, one commenter experienced
- The beachcomber author said Claude argued with changes because it read old docs.
- They are experimenting with canonical specs cascading canon, code, docs.
- LLMs were weak at 3D reasoning, one commenter noted
- They had an LLM write a cadquery interpreter for CAD and found 3D reasoning weak.
- Another commenter said AI removes friction but writing code by hand remains satisfying.
How to Apply
- Run your agent in a sandbox
- Today, run Claude Code or similar inside a podman container or separate OS account.
- Give it a copy of the workdir and pull changes back only via git diff.
- Constrain LLM output to a JSON contract
- Don't let the LLM write config files directly; have it fill a schema-bound JSON.
- Convert only validated values to XML, Terraform, etc.
- Define a canonical order for agent docs
- State in the repo which wins: spec, then code, then docs.
- Write a rule so the agent doesn't change either arbitrarily when they disagree.
Try It
# Paste into Claude or ChatGPT (simplified version of the JSON-contract approach from a comment)
You are a config generator. Output only JSON for the request below. No explanation.
Schema: {"name": string, "params": {"{{param1}}": number, "{{param2}}": number}}
Value range: 0-10
Request: {{describe the config you want}}Terminology
Related Papers
Rust off Rails: The opinionated web stack for the AI era
Sixty Markdown files tell coding agents how to build a Rust web app, with no framework code.
LayerFS: The filesystem for AI coding agents
Mount an ephemeral workspace per agent, commit only the changes
Launch HN: Alter (YC S25) – Access control for multi-user agents
Alter gives each agent its own identity, policy, and audit trail instead of shared API keys.
Using Claude to reduce the attack surface of code
Claude scanned 25 open-source projects and found 493 needlessly reachable code issues.
WhatsApp's help pages hid LLM instructions: dont recommend a regulator
WhatsApp help pages hid 102 AI-only instructions, including not recommending a data regulator
Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy
Saving long tool output to files cut GPT-5.1 accuracy by 11.3 points; naming the authoritative file restored it.
Fake News, Real Crashes: Information Injection Attacks on LLM Financial Trading Agents
Fake headlines hijacked LLM trading agents 62% of the time versus 14% for prompt injection.
Harness Engineering 101: The LLM is the brain, the harness is the body
An agent is a loop managing a JSON array; 16 chapters build one from scratch