Agentes de IA reservaron cuatro Uber.

OpenAI has agreed to acquire Pierre Computer, a code-hosting startup. The move isn't exactly random; Pierre Computer already hosts the code repositories for Codex, giving OpenAI an existing operational tie to the platform.

This acquisition puts OpenAI on a direct collision course with GitHub. By bringing repository hosting in-house, OpenAI is signaling a strategic shift away from just providing the models that write code, moving toward controlling the infrastructure where that code actually lives and breathes.

For developers, this means the AI tooling wars are moving down the stack. OpenAI doesn't just want to be your copilot anymore; they want to be the entire plane. Expect tighter, proprietary integrations that make leaving the OpenAI ecosystem increasingly difficult.

Meta has acquired Conductor, a startup that built an interface specifically designed to direct multiple coding agents simultaneously. As part of the deal, Conductor’s existing hosted service will officially shut down on November 20.

The acquisition points directly to Meta’s growing obsession with multi-agent AI systems for software development. Managing one AI coding assistant is hard enough; orchestrating a fleet of them requires specialized infrastructure that Meta clearly decided was cheaper to buy than to build from scratch.

This is a massive validation for the multi-agent architectural approach. Builders should take note: the major labs are no longer just looking at single-prompt interfaces. The future of AI development tools is orchestration, and Meta just bought a conductor to lead the band.

Nothing CEO Carl Pei has announced that the company’s first AI-native product will ship this December. The catch? The new device will not feature a screen, teasing a completely new, screen-less interaction model for users.

It is a massive gamble. The industry has already watched high-profile screen-less AI hardware stumble out of the gate, proving that consumers are deeply attached to visual feedback. Pei is betting that Nothing has solved the interaction puzzle that others missed, challenging our absolute reliance on displays.

For hardware builders, December will be a fascinating case study. If Nothing succeeds, it opens the door for a new wave of ambient computing devices. If it fails, it might permanently kill the dream of the screen-less AI wearable.

Moonshot has officially released the open weights for its Kimi K3.1 Pro model. The release includes the ability to read both text and images, and most notably, it boasts a massive 1-million-token memory capacity, all governed under a custom license.

This is a heavy swing in the open-weight ecosystem. While other labs are busy optimizing smaller models, Moonshot is democratizing access to a genuinely massive context window, proving that multimodal, long-context AI isn't just for proprietary API gatekeepers anymore.

Open-source builders just got handed a serious piece of artillery. If you are building applications that require digesting entire codebases or massive document troves, Kimi K3.1 Pro is your new baseline.