Explore what's new in AI

OpenAI Devs showcases community-built Codex projects

The project roundup provides concrete examples of what developers are building with Codex, with linked implementations that can inspire reproducible experiments.

Moshe Kanat frames disposable and durable software as a spectrum

As experiments become persistent tools, reducing the cost of engineering rigor can help teams improve reliability without overbuilding every prototype.

Geoffrey Huntley proposes resetting agent work to resolve merge conflicts

Instead of manually merging parallel edits, the proposal resets the working copy and asks the agent with context to reapply changes, simplifying coordination at the cost of rework.

A code-review flowchart puts agent checks before human escalation

The proposed loop adds self-verification, a second agent, and automated checks before shipping, reserving human attention for novel issues rather than routine failures.

An Xcode cleanup prompt recovers about 420 GB from runtimes and caches

A size-aware keep-or-delete plan reclaims hundreds of gigabytes from old simulators and build data while requiring approval before removing files.

GitHub Copilot maps saved prompts and sessions to Logitech MX Keypad

Physical keys can trigger prompts, resume sessions, or run multi-step actions, providing configurable shortcuts for repetitive agent interactions.

Sphinx quizzes developers on agent-written diffs before Git pushes

A pre-push hook tests whether developers understand a change before it ships, adding a lightweight human-verification step to agent-assisted coding.

OpenAI adds editing of earlier conversation prompts

Users can revise an earlier prompt and continue from that point, discarding later responses and tool calls to make conversational iteration easier.

A developer alleges Apple is restricting app feature switches

The claim raises a possible constraint on app rollout and testing patterns, though the post does not provide enough detail to establish the scope of Apple's policy.

Theo demonstrates remote session navigation in a T3 Code nightly build

The demo shows sessions running across VPSs and Linux machines, illustrating how remote execution can fit into a coding-agent workflow.

Theo reports many T3 Code Codex users also use Claude Code

The reported overlap suggests developers may mix coding assistants rather than commit to a single agent, although the figure covers T3 Code users only.

Perplexity Computer builds a playable browser driving game

The demo shows an agent producing a navigable driving game in a browser, offering a concrete example of computer-use systems generating interactive software.

A practitioner demonstrates Tern's integrated agent workspace

The workspace combines Git, browser, database, task management and collaboration features, illustrating an attempt to keep agent development inside one environment.

Martin Odersky discusses capability-oriented programming in Scala

The keynote discussion connects capability-oriented programming with language design choices, offering a reference for engineers evaluating effects and resource boundaries.

Uncle Bob recommends instrumenting bugs with an agent iteratively

The workflow asks an agent to write tests and add instrumentation before investigating a bug, shifting debugging effort toward providing a reproducible harness and context.

Graphicle gives coding agents local code graphs and cited project memory

The Rust tool combines repository structure, documentation, and prior sessions to return cited project context without requiring a remote indexing service.