Nathaniel Watson

Plan intentionally. Test systematically. Evaluate honestly. Wise words from Danielle Maxwell in her latest blog post on how to effectively use AI to build solutions. AI gives us the ability to solve more complex problems, but building high-quality applications still requires a focus on the fundamentals. If you have struggled with prompt engineering for your specific use cases, check out Danielle’s blog post for tips on getting more reliable output from your models. https://lnkd.in/esipXaYC

15

AI meeting features aren’t a “nice-to-have” anymore. 68% of SaaS users said they’d switch products for them (State of Meetings Report, 2025). On Nov 13, Hazik + Nick are walking through how the Notetaker API helps teams deliver these features quickly — without spending months stitching together email/calendar integrations and transcription services. For developers, the interesting part is how they’ve simplified the pipeline so you can focus on building value on top of the data, not just moving bytes around. Register Here 👉https://lnkd.in/ejMhp5kb

46

As more of our output is generated, modified, and maintained by AI agents, verifiability becomes imperative. Not just at the task level: “Did the agent complete the task correctly?” But at the system level: “Can we still understand, audit, reproduce, and trust the system as agent-generated changes compound over time?” Without verifiability, agentic AI will not just create productivity gains. It will create extreme maintenance risk: systems full of automated decisions, outputs, and modifications that no one can confidently validate or trace back to intent, requirements, and constraints. Which will grow over time as agent output compounds, until a system ends up entirely unruly and unmanageable. The real unlock is not just autonomous agents and orchestrated workflows. It is autonomous systems that remain verifiable as the product scales.

22

2 Comments

Geoff Charles at Ramp just published one of the more eye opening AI adoption peices I've seen. Non-engineers making thousands of production code contributions per month. 1,500+ internal apps shipped in six weeks. https://lnkd.in/e6CMB2Tr Worth a full read if you're thinking about how to actually move your org on this stuff. They didn't have a master plan. They had a culture that rewards speed, and they kept doubling down on whatever was working. "Training doesn't work. Office hours and workshops help. But the world's best teacher is staring right in front of you: it's AI. You can only lead a horse to water. The single biggest unlock is getting someone to experience a real result on day one." "The number one way companies kill AI adoption is by treating it like a procurement decision. Budget approvals. IT reviews. Token limits."

22

My latest post about how I am actually using AI to ship production software has been published on Medium! In it I will walk you through the tools I'm using, and my process for iterating on production applications. https://lnkd.in/gPJ7k8TK

12

Great tips here. It takes practice, but it’s so freeing to just let the little things go. I find the mindful acceptance and reframing techniques especially helpful. “We’re living in such very stressful times—antagonistic, hostile times—and people are under a lot of stress... We need to try to think of ways, even just in your own network, to help people feel better.”

7

1 Comment

A lot of organizations' entire AI strategy is, basically, "make more documents, faster." It's a very natural entry point. Feed in a transcript, get back clean notes. Use a note taking app, get clean notes, store on shared drive. Instant, and it makes everyone feel like a genius. Great first use of the tools. The problem is a lot of them never leave this stage. Every meeting becomes a summary, every summary a doc, until someone's building a system (a new knowledge base!) to organize the docs nobody reads. This is my nightmare. The tools got really good at making documentation. They did nothing to make it worth reading. That part is still on us. I wrote about that this week 👇 https://lnkd.in/gi7a-KUG

21

7 Comments

Spent years at previous companies building and maintaining integrations. It's brutal. At Payflow, we use Merge. Their observability tooling (logs, real-time alerts) makes debugging sync errors trivial. FP&A teams need data they can trust. Merge helps us deliver that. Full case study: https://hubs.li/Q03-v50v0

47

10 Comments

"Rows, Columns, and Consequences" might not be the biggest or most famous conference we've spoken at, but it's certainly got the best name. Shomik Ghose was there last month talking about building Pangolin, a production database that 3 engineers shipped in just 14 months -- because they were testing with Antithesis. https://lnkd.in/ewBHU58N

39