
Danny G
@the_danny_g
Followers
1K
Following
15K
Media
210
Statuses
2K
AI Founder | CTO @LegionHealth (YC S21) | ex-MSFT PM
Austin, TX
Joined September 2014
This is something missing in the Full Stack AI narrative. In full stack AI, the company isn’t agent-ready or AI-enabled. The company itself is fundamentally designed for AI from the ground up. Every hire, process, SaaS tool, and product is built with these tradeoffs in mind.
Context engineering is increasingly the most critical component for building effective AI Agents in the enterprise right now. This will ultimately be the long pole in the tent for AI Agents adoption in most organizations. We need AI Agents that can deeply understand the context.
1
1
1
@dakshgup @greptileai “Notion AI has gotten really good. We were prepping an annual review for a customer and it pulled quotes, video clip moments, key stats and graphs—things that would have taken me hours.”. (Wei-Wei Wu—CEO @ Momentic + @NotionHQ).
1
0
2
@dakshgup @greptileai Which AI tools do you pay for and get the most out of?. “Claude Code. It’s all in how you use it. DB schema issue came up, an engineer used Claude Code to spin up a UI and entire internal tool for that one issue to help troubleshoot. Issue got fixed, and that UI was deleted five.
1
0
1
@dakshgup @greptileai “We’re still early into how we abstract our tools and workflows. As we’ve learned more, it’s easier to simplify and build even faster”. (Wei-Wei Wu—CEO @ Momentic).
1
0
1
@dakshgup @greptileai Recent underrated developments in AI. “Fine tuning is viable—maybe for the first time. Synthetic data has gotten a lot cheaper, as have open source models.”. (Ben Hylalk—CTO @ Raindrop). “Tool use. Paradigm shift in the past 3–4 months. The correct way to build agents is now.
2
0
1
@dakshgup @greptileai “The difficulty of writing good software has gone down massively”. “Taste REALLY matters as you’re building these projects.“. “We’re heading in the direction we believe in SO much faster [(as are others)]. That makes the direction matter so much more.”. (Raunak Chowdhuri—CTO @.
1
0
1
@dakshgup @greptileai “It’s really telling to look at OpenAI—increasingly you see them taking verticalized bets on products. The growth is coming not from percentage point gains on intelligence alone; rather, increasingly, it’s vertical applications”. “Open source models are getting to the point where.
1
0
1
@dakshgup @greptileai “Choosing a good data model is the MOST important thing” (Ben Hylalk—Raindrop). - On Clickhouse, notoriously hard to migrate, similar for other things across stacks ➡️ need to nail it. ➡️ Don’t vibe code your data model.
1
0
1
“Startups are structurally better equipped [than enterprise] to build great AI software” (@dakshgup @greptileai). - Enterprise has established risk mitigation processes.- These processes assume deterministic software.- Enterprise values determinism / downside protection over 🚀.
1
0
1