Agentic Engineering · 2026-10-01
AI engineers need to move beyond just coding to "shape the build" by mastering four key skills. These include actively driving the build loop through prototyping and iteration, making sound product decisions even with incomplete specs, enhancing communication and leadership to translate technical concepts across diverse teams, and demonstrating high agency and ownership to proactively solve problems and guide direction.
AI Honeycove · 2026-10-01
Glasser is an AI agent skill that allows language models like Claude, Cursor, or OpenAI Codex to access premium data from multiple providers without requiring separate subscriptions or manual workflows. By integrating Glasser, users can make data requests directly within their AI agent, specifying needs like company research, lead generation, SEO keywords, or social data. Glasser automatically finds a suitable data endpoint, displays the provider and exact price before execution, and only charges for the data actually consumed, streamlining data acquisition and reducing costs.
AI Engineer · 2026-09-30
AI Engineer · 2026-09-30
Agentic Engineering · 2026-09-29
This video discusses a sensible AI writing policy from Clay, emphasizing that AI should reduce work, not export it to readers. The core principles include owning every sentence you write (even if AI helped), recognizing that writing is thinking and should lead to deeper understanding, and valuing brevity over length. The goal is to leverage AI to make content clear and concise, providing a one-time cost for the author that benefits all readers.
Theo - t3.gg · 2026-09-22
Beyond Coding · 2026-09-16
Theo - t3.gg · 2026-08-20
Simon Willison · 2026-10-01 · 1 min read
AI agents can propagate worm-like attacks by embedding malicious instructions in shared communication channels (email, Slack, documents), causing one compromised agent to infect others it interacts with. This matters because sandboxing alone is insufficient to contain rogue agent behavior when agents share any common medium, meaning any multi-agent system you build today likely has this attack surface regardless of how well you've isolated execution environments.
Beyond Coding · 2026-06-24