Explore what's new in AI

Restream supports custom scenes and multiple hosts for livestream production

A practitioner chose Restream for multi-host streams, shared screen content, and multistream output, useful capabilities for live technical demos.

Geoffrey Huntley argues modern tooling makes unikernels practical again

Huntley revisits unikernels as application-and-OS bundles, pointing to smaller attack surfaces and better tooling as reasons to reconsider them.

Zeeg argues agents will not erase the complexity of professional software

The thread distinguishes simple task automation from dependable production software, arguing domain expertise and operational risks still matter when agents write code.

Dzhng uses custom validators to guide long-running coding agents

Custom performance and usability validators help agents build Rust and Wasm apps while checking requirements during development.

Hermes Agent consolidates official plugins in a monorepo

Nous is consolidating official Hermes plugins to centralize discovery and issue tracking, while clarifying the repository covers Nous-maintained plugins.

Zeeg releases a Windows helper for separate Claude profiles

The helper runs work and personal Claude profiles in parallel with separate credentials and distinct window colors.

Theo recommends T3 Code's nightly build after recent improvements

Theo asks former users to retry T3 Code's nightly build, citing improvements without detailing the changes.

Theo rejects paid-product integrations in T3 Code

Theo says integrations should serve users rather than function as distribution channels, warning that paid-product tie-ins can bloat an app.

Theo explains why he built the ts-rust compiler

Theo shares a video explaining the motivation behind ts-rust, a project with goals beyond faster type checking.

John Rush says Claude Code built a table-simulation wizard in five minutes

Rush reports generating a step-by-step simulation wizard in minutes, illustrating how coding agents can accelerate bespoke tools for physical projects.

Kuncheng Guid measures maintainability by the cost of future changes

A codebase that takes far more tokens to modify can impose real costs even if its current output works, making future change effort part of maintainability.

Theo recommends Linux to avoid environment friction with coding agents

Theo argues that a conventional Linux setup can reduce friction when coding agents interact with local development environments.

T3 Code adds a visual callout for its goal workflow

The nightly build adds a /goal flow that keeps a long-running objective visible while an agent works, making open-ended runs easier to steer.

OpenCode prototype renders interactive interfaces inside the terminal

A local prototype shows interactive interfaces rendered inline in a terminal, an alternative to switching between agent chat and external tools.

Theo moves ts-rust builds and tests to Blacksmith runners

Moving builds and tests to remote runners reduces local bottlenecks during parallel development, though individual runs still take time.

Madiator alleges Perplexity uses Nimble code without attribution

Madiator alleges Perplexity used Nimble code without attribution and asks for a remedy; the post does not include a response or license analysis.

Matt Pocock revisits the maintenance cost of tactical tornadoes

Developers can ship quickly while leaving cleanup for teammates; agent throughput makes it important to measure maintainability and downstream costs, not output volume alone.

Wi-Fi 7 chipset support can require a newer Linux kernel

A developer reports recent Wi-Fi 7 hardware failing on a supported Ubuntu kernel and working after a manual upgrade, a compatibility wrinkle for workstation builds.