The Radar
Tuesday, 23 June 2026
Today's picks
Ponytrail
AI CodingLocal audit trail for AI coding agent edits.
As AI coding agents touch more of your codebase, knowing exactly what changed and when becomes critical. Ponytrail runs locally and logs every edit an agent makes, giving you a proper paper trail without sending anything to a third party. Simple idea, serious utility.
Selector Forge
Developer ToolsBrowser extension that generates resilient AI-powered CSS selectors.
Brittle selectors are one of the biggest pain points in browser automation and testing. Selector Forge uses AI to generate selectors that survive DOM changes, which matters a lot if you are building agents that scrape or interact with live pages. Worth keeping an eye on for anyone doing web automation.
Also on the radar
PMB
AI CodingPersistent memory for coding agents is still an unsolved problem for most teams. PMB tackles it locally over MCP, which keeps your context private and avoids cloud round-trips. Early days but the approach is sound.
Ask a Human
AI AgentsFully autonomous agent loops still get stuck, and most teams have no clean way to escalate. This tool pages a human when the agent hits a decision it cannot make alone, keeping loops unblocked without requiring constant babysitting. Practical and honest about where AI still falls short.
Keynap
Developer ToolsOn-device voice typing means no audio leaving your machine, which is the right default for anything sensitive. Keynap runs the whole pipeline locally on Windows without a subscription or cloud dependency. Straightforward and worth a look if you dictate a lot.
OpenPlan
AI AgentsThe Waze framing is a bit marketing-speak, but the underlying idea is interesting: dynamic replanning for agents as conditions change mid-task. If it delivers on that, it fills a genuine gap in multi-step agent orchestration. Worth watching as the agent tooling space matures.
Hacker News
Show HN: Selector Forge – browser extension for AI-generated resilient selectors
33 pts 1 commentsA browser extension from Intuned that uses AI to generate CSS selectors that hold up when the DOM changes. Useful for anyone building web automation or browser-based agents. Direct answer to one of the most common causes of flaky test suites.
I built Ponytrail, a local audit trail for AI coding-agent edits
23 pts 11 commentsPonytrail logs every file change made by an AI coding agent, locally, so you have a full record of what was touched and when. Built in response to the real problem of agents making opaque edits across a codebase. Gets more relevant as agentic coding becomes the default.
GLM-5.2 is the step change for open agents
9 pts 0 commentsGLM-5.2 from Zhipu AI is being positioned as a strong open model for agentic tasks, with independent evaluations placing it above GPT-5.5 on knowledge work benchmarks. The interconnects.ai write-up goes deep on why the architecture suits multi-step reasoning. Good news for teams wanting capable open models without API lock-in.
Show HN: PMB – local-first memory for AI coding agents over MCP
7 pts 6 commentsPMB gives AI coding agents persistent, local memory via the Model Context Protocol. It keeps context on your machine rather than in a hosted service. Early project but the local-first approach is the right instinct for developer tooling.
Show HN: A private pager for your AI agent loops
3 pts 1 commentsAsk a Human lets you define points in an agent loop where a human needs to approve or unblock the next step. It sends a notification and waits for a response before continuing. Honest about the limits of full autonomy and genuinely useful for production agent workflows.