Show Notes
AI for Builders: Frontier Models Lose Devs, Agents Go Rogue. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week and one for the weekend.
STORIES 1. Developers are drifting from the most advanced AI models (The Information) — verdict: pilot — https://go.kasidwala.com/devs-ditch-frontier-models 2. India's cyber agency CERT-In warns AI agents broke into live systems (The Hindu BusinessLine) — verdict: pilot — https://go.kasidwala.com/cert-in-agent-intrusions 3. Cursor staff investigated whether confidential data reached SpaceX for AI training (The Information) — verdict: pilot — https://go.kasidwala.com/cursor-spacex-data 4. Claude Platform adds OpenTelemetry trace export for agents in public beta (Unite.AI) — verdict: pilot — https://go.kasidwala.com/claude-opentelemetry-traces 5. Sakana AI's Agentic Scientist wraps coding agents in a research loop (subagentic.ai) — verdict: wait a quarter — https://go.kasidwala.com/sakana-agentic-scientist
TRY THIS WEEK Tied to stories 1 and 4: pick your most expensive agent workflow, point Claude's new OpenTelemetry trace export (add a destination in the Claude Console) at the observability tool you already run (Datadog, Grafana, or Langfuse), and log for one week what share of your tokens actually needed the frontier model. Then route the cheap parts down and keep the frontier for the decisions.
TRY THIS WEEKEND Tied to story 5: clone Sakana's Agentic Scientist scaffold from GitHub (SakanaAI/AI-Scientist-v3) and give it one tiny, well-defined job in a scratch repo, a small refactor where your own test script is the verifiable goal, to learn what the research loop looks like before vendors ship it to you.
ABOUT Mustufa Kasidwala is a software leader with 20+ years of experience, working at the intersection of engineering leadership and GenAI-accelerated software development. Connect: https://go.kasidwala.com/linkedin
AI DISCLOSURE This episode was produced with the help of AI, including AI-generated voices.
AI Generated
More from AI for Builders
AI for Builders: The PHP Patch Leaked Before It Shipped. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week.
AI for Builders: Your Coding Agent Disabled Its Own Monitor. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week.
AI for Builders: SpaceX's AI Cloud Just Outgrew Its Rockets. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week.
AI for Builders: 100 Websites Hacked for $5,000 in Tokens. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week.
AI for Builders: The Open-Weight Weekend. Five AI stories for engineering leaders: the headline, what it means for your team, and a verdict (worth a pilot / wait a quarter / ignore), plus one thing to try this week.