Your team culture matters more than your AI tool stack
Two talks converge on the same hard truth: psychological safety, not model choice, is what makes or breaks AI adoption.
The third shows why the file format itself is suddenly a cost lever for GPU workloads.
Safety over tools. Erin Doyle argues AI adoption fails or succeeds based on whether engineers feel safe admitting ignorance, not which tools they pick.
Teams that skip the culture conversation get polarized adoption and widening workflow gaps.
Agent governance now. Claude Code's new mod platform turns agents into mutable software with hooks, sub-agents, and custom selectors.
The feature is a governance and security challenge, not just a productivity win: every mod adds surface area for prompt injection and data exfiltration.
GPU idle tax. Vortex claims 30x faster GPU scans than Parquet by cutting CPU decompression and memory copies.
The bottleneck diagnosis is credible but the numbers are vendor-run, making this an evaluation trigger for GPU-heavy teams, not a migration mandate.
My take
I think the safety-first argument is the one that actually pays off, but it's also the hardest to execute on because you can't buy it.
The Vortex pitch is the kind of concrete infra bet that an engineering manager can actually evaluate in a sprint: convert one table, measure the real gain.
Mods are the wildcard. Every team should audit their agents now, not when the first incident hits.
The three videos share a thread: the hard problems are people, governance, and data plumbing, not model benchmarks.
The videos behind this brief
We watched these so you don't have to.
Your AI tool choice matters less than your team's culture of safetyInfoQWatchThe team conversation framework is immediately actionable in your next retro.
Claude Code Mods: The Platform Shift Most Teams Aren't Prepared ForLatent Space
Parquet is bottlenecking GPU pipelines: Vortex claims 30x faster scansInfoQVideo and thumbnails remain the property of their respective creators. tldw.news provides editorial analysis, commentary, and discovery links to original content.