Jim Speth

I had an absolute blast chatting with Tracy Lee and Danny Thompson on the Leadership Exchange Podcast about what it really takes to build intelligent systems in the AI era. We went deep on: 🤖 The shift from single-model thinking to agentic ecosystems 🧭 Why judgment and evaluation matter more than ever 💼 How AI and Product roles are evolving 🧩 Why data quality and governance quietly make or break AI success 🎯 What the rise of the Chief AI Officer tells us about the future of leadership 🔥 Hot take: AI transformation is 20% tech, 80% culture. If you’ve ever wondered how to evolve your teams, systems, or career in this new AI landscape — this one’s for you. 🎧 Watch here → https://lnkd.in/gABmqvTb What’s your biggest challenge in scaling AI beyond prototypes? #AI #Leadership #Innovation #ProductManagement #DataScience #AIAgents #Engineering

33

2 Comments

I keep seeing a number of people today who say they're getting garbage out of their agent, or that the code is brittle, or, or, or. The secret isn't the prompt, it's the workflow. How many of the following do you have answers for? * Does the agent really know what you want? How do you know? * Do you know what the agent is going to do? * How can the agent verify it's work? * How does it know what the structure of the system it's adding code to? That is, what's often the cultural memory of the builders. * How much are you guessing at what the agent can figure out? The answer is having workflow to address this. I've found starting prompts can be extremely terse, vague, incomplete, and so on, and /still/ get good outcomes if you have good workflow.

7

1 Comment

Context as the new code, I like this framing, a ton. So much I’m orienting some new builds around the context flywheel and delivery lifecycles. That said, if context is the new code, we have a shit load of tooling to build.

8

3 Comments

Excited to share a post I co-authored with Rong Hu on our large-scale migration to an in-house observability platform. We cover key lessons learned that we hope are useful for anyone tackling a similar challenge. https://lnkd.in/e6jfvWWv

78

3 Comments

🚀 Figma to Qt 1.0 is live and I'm incredibly proud to have led the engineering effort that made it real. I worked on this one for a while, and the problem it solves is one that's been frustrating embedded UI teams for years, the gap between what a designer signs off on in Figma and what actually ends up on the device. You preview your GUI live as running QML in the browser, catch issues before they become expensive, and export QML code - no ambiguity, no reinterpretation on the developer's end. It's a deterministic converter, not a generative tool. That distinction matters more than it sounds when you're shipping to a medical device or an automotive HMI. For teams building embedded UIs in automotive, medtech, industrial, or consumer electronics — this changes the designer-to-developer handoff entirely. https://lnkd.in/gqByXgyP https://lnkd.in/giFijphZ #Qt #FigmaToQt #EmbeddedUI #QML #HMI #Figma

37

2 Comments

This is an older talk, but one I refer to often and recommend to current tech leads and those on a path to becoming a tech lead. Eryn summarizes the shift from being a full-time coder to balancing technical oversight with advocacy, facilitation, and motivation - ensuring that everyone else has what they need to succeed. Technical leads are the bridge between the business stakeholders' goals and the engineering team's ability to deliver. Definitely worth a watch (on repeat)! #engineering #leadership #techlead #leaddev https://lnkd.in/gDypDS78

12

Clay doubling down on increasing action costs while a lot of gtm-e world is laser focused on BYOK workflows with Claude Code is an interesting tension. In the last ~week Cargo shipped a CLI, Apollo.io released an MCP connector for Claude & most of my feed is Claude Code. The ecosystem energy feels entirely focused on terminal-first. I’ve loved Clay for orchestration but have increasingly felt that the UI/UX is slowing me down as I ramp up my terminal-first workflows. That said, terminal-first isn’t multiplayer yet and isn’t in it's final form - esp for larger companies where multiple humans benefit from a shared UI. What’s interesting is this is coming from the people who coined “GTM Engineer” in the first place. The early adopters Clay built the category for are the ones feeling the pull away from the orchestration layer Clay monetizes. The dispersion is probably temporary. But if Clay’s pricing move take them further from the early adopters who built the category, that’s a deliberate bet to cross the chasm to enterprise, while some subset of builders build elsewhere. Interesting to watch.

17

3 Comments

I had a crazy idea: what if recommenders stopped being “similarity engines” and became relationship engines? GraphRAG + rec systems sounds wild… until you try it on real estate: commute • HOA rules • school zones • neighborhood vibe • “quiet but walkable” preferences I wrote a breakdown on when GraphRAG helps, how to build it without overengineering, and when it’s a complete waste of time. Read it here 👇 🔗 https://lnkd.in/gqfUqrFJ #GraphRAG #RAG #RecommendationSystems #AppliedAI #LLM #MachineLearning #AIEngineering

32

1 Comment

A lot shifted in 2025. AI made some things easier and other things more complicated, and it pushed teams to focus more on resilience. The 2025 State of the Internet Year in Review breaks down what stood out most this year. https://lnkd.in/eh2eZXfA

14

Ah! I didn't expect to see this in my feed today, thank you OfficeSpace Software for featuring me today. This interview was recorded a few months ago, and a lot has changed in the world of AI-driven engineering since then, but I still wouldn't change a thing about what I said back then. Be an architect, not a vibe-coder. Build plans and review them thoroughly before allowing AI to execute them. And DO NOT let AI over-engineer your plan, you don't want to end up having to maintain things that are never used in production, or worse, having to respond to a SEV-1 caused by one of these feature that was not part of the spec, but sounded cool to add at the time.

15

I am excited to share with you the first post of our Engineering Blog! In this issue we cover some of the key engineering decisions we made that have enabled our lean team to build a multi-cloud, regional platform that is used by enterprises today. We will be sharing deeper dives into many of the topics mentioned in this post. If the problems and the tech we are building at Thread AI excite you, we'd love to hear from you!

48

2 Comments

In the world of Retrieval-Augmented Generation (RAG), pure vector retrieval often suffers from reduced recall when dealing with sparse queries or out-of-distribution terms. That’s where hybrid search comes in. By combining dense vector retrieval with lexical methods like BM25 (TF-IDF + document length normalization), we can significantly improve recall and overall answer accuracy — especially in production-grade agent pipelines. Don’t just chase semantic similarity — anchor it with lexical precision. #HybridSearch #RAG #BM25 #InformationRetrieval #LLM #VectorSearch #AIEngineering #MIR #SearchSystems

14

Scaling from 0 to AI: A year of breaking the 10x barrier. One year ago at Hertz, we hit a classic scaling wall. As we pushed to transform our customer experience, our manual QA process became the anchor holding us back. In a legacy environment, the standard move is to hire more people or buy more "enterprise" tools. We took the unconventional path with a focused team and a relentless drive for agentic innovation. The Mission: Transform the Culture, then the Tech. We didn't just want to automate; we wanted to democratize quality. That meant a radical shift in mindset, skills, and how we defined "done." The Hustle: - The Pilot: I started with a hands-on POC to see if GenAI could actually handle the messiness of real-world UI workflows. - The Prototype: We didn't wait for a perfect roadmap. We built a production-grade prototype to prove that "Agentic" testing wasn't just a buzzword—it was a force multiplier. - The Buy-in: I pitched leadership on a high-stakes partnership with AWS to bring this to life at scale. The Result: With Amazon Web Services (AWS) Nova Act now in General Availability, we’ve compressed weeks of manual toil into minutes of AI-driven precision. We’ve effectively turned our QA org into an innovation engine that builds at startup speed. Check out my full testimonial on how we made it happen: https://lnkd.in/gHzsiAS5 The Vision for 2026: This wasn't just about testing. It was a blueprint for how "Agentic" workflows can gut-renovate legacy processes. My aim for the future is to keep finding these high-friction areas and using cutting-edge tech to blow the doors off. If you’re interested in finding out more about how to put agentic workflows into production, let’s talk. #StartupMindset #EngineeringLeadership #AI #DigitalTransformation #AWS #NovaAct #HertzTech #InnovationAtScale

92

8 Comments

Next week I’m speaking at LeadDev LDX3 on the Ways of Working stage — all about team operating systems: what they are, why they matter, and how to build a great one! Hope to see you there. And to get in the zone, here’s a quick primer on systems thinking more generally. Huge thanks to GC.NYC #girlcrew - Woman-Owned Video Production — a phenomenal woman-owned video production studio — for bringing this video to life. Emily Geraghty and her crew are magic. If you’ve got an idea, reach out for help making it real!

82

10 Comments

iOS build submitted for App Store review today. Built solo over 24 working days: 347 commits, 70 unit tests, 8 Maestro automated E2E test flows, and ~25k lines of TypeScript across 15 feature modules. Powered by Expo and Tailwind. Android next.

45

4 Comments

Telling my teams (and myself) the simultaneous importance of: - running multiple sessions/workstreams in parallel to take full advantage of AI tooling *where possible* - NOT running so many that you lose focus/concentration/diminishing returns kick in Better to ship the things you can with full focus and attention instead of twice as much broken code. The line between sloppy and 10x engineer can be thin sometimes. Don't be a slop-cannon.

14

5 Comments

rqlite -- and the way I develop it -- is fundamentally changing. LLMs are in the loop -- from ideation to Pull Request. https://lnkd.in/ejwXGddR

27

1 Comment