Загрузка видео...
Не удалось загрузить видео
Code generation isn't the moat. Engineering is. Last week I sat on a panel at the SVB Experience Center with Sanjeev Dhanda (Google DeepMind) and Abhi Aiyer (Mastra) on the move from vibe coding to agentic engineering Constructive Silicon Valley Bank Kristopher Floyd
971,982 просмотров • 4 месяцев назад •via X (Twitter)
Комментарии: 20

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd Good engineers, good cities, people are comfortable, people live easy and safe lives.

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd If 41% is AI-generated, trust layer becomes the real product, not the feature anymore

the part that breaks is when state needs to span more than two files. one-shot generation handles screen one fine. screen two needs a model someone has to reason about, and that's where the engineering either shows up or doesn't. we built mk0r to enforce that cross-file state model as a first-class graph the agent has to reason against,

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd agentic engineering still fails quietly without observability, evals and rollback baked in

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd Awesome

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd The industry’s realizing that shipping real products takes more than AI-generated code. Tools can accelerate coding, but architecture, reliability, debugging, product judgment, and systems thinking are still what separate demos from actual engineering.

41% of all code is now AI-generated. The question isn't "can we build it?" anymore — it's "can we trust it?" Coding is being commoditized. Engineering is becoming the discipline of trust.

You don't vibe code the vault. The durable value lives lower in the stack: authorization, testing, observability — the substrate that makes agentic software safe at scale.

Thanks to @Kristopherfloyd / Frontier Syndicate and @SiliconVlyBank for hosting 🙏🏻

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd Engineering’s true strength lies in the vision and discipline behind the code, not just the code itself. It’s the craft that really differentiates.

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd Exactly. AI raises the floor for code generation, but architecture, tests, permissions, and operational judgment are where real engineering still compounds.

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd what you're looking for 👇

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd what does "engineering" look like in practice here, graphs, eval harnesses, deploy pipes?

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd engineering is building of systems that help humans and agent converge on reliable software, opposed to just generating code

@GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd

@iamhustler30 @GoogleDeepMind @abhiaiyer @mastra @constructive_io @SiliconVlyBank @Kristopherfloyd
