Craig Lee

How I built a small business app for my wife's family Tl;dr, 1) Use Claude Artifacts to build basic functionality 2) Use cursor / claude code to build it out fully 3) Make mistakes and refactor to different technologies more than I probably should. Long version: 1) Use Claude artifacts to make a simple form 2) port it into GitHub using create react app, connect to vercel (super easy!) 3) get it working (can create and share receipt images, woohoo! In laws love it) 4) move to vite (apparently CRA is old and AI used it) 5) move to nextjs to have a semi backend 6) convert js to typescript (that took awhile) 6.1) Revert at least once due to urgently needed new features. 7) Add a database using Neon 8) Decide not use to Neon, use Prisma instead (because AI suggested that) 9) decide not to use Prisma, use railway instead 9.1) Also switch from Prisma ORM to postgres.js 10) fight with AI not to use "unsafe" 11) build different data views and UI for all database operations (add, edit, delete) 12) Add multiple tables and build out more advanced functionality (like expenses!) 13) learn how to backup and migrate databases 14) create a dev and prod instance in GitHub + railway, start pushing to dev for testing instead of prod 15) eventually try to make a pdf report, fail, build f# backend to make pdf 16) Learn type algebra to and build a super elegant user permission management system. 16) Start moving APIs to f#, make elegant schema system 16.1) Break UX, then fix it. Users slightly confused but they're my in-laws so they'll be ok. 17) Accidentally wipe my local database (it was my fault, not AI's), then restore it from a backup. You are here Next steps  - build ability to dynamically build forms. - find another small client - teach AI to build with me Lessons : I find that AI doesn't create bad code, it just has generic assumptions. As I guide it to have better assumptions it gets me to something that I find elegant and easy to work with. Still it starts messy and I have to massage it to something clean. I also find that I move faster if I think first, but often I like to skip that and do first. This loses time (note to self).

5

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

Perfectly (YC W26) has the most precise talent sourcing engine on the market. Our sourcing system scored 96.9% precision on Exa's open-source People Search Benchmark — evaluated on engineer and sales roles across SF/Bay Area & NY Metro. For comparison: Juicebox scored 81.5%. Exa scored 62.3%. Most sourcing systems run keyword retrieval — match on title, location, maybe a skill tag. We built a recommendation system (our team comes from TikTok and Meta) that models both sides: the hiring context and the candidate signal, then ranks by predicted fit. We've put this to work with Giga, LlamaIndex, Corgi, Porter, and Metaview — filling their engineering and sales roles with candidates who actually pass interviews. But sourcing is only one stage of the pipeline. Perfectly is a full AI-native recruiting agency — intake, strategy generation, multi-channel sourcing, automated outreach, screening — delivering interview-ready candidates within 24 hours. Backed by Y Combinator. Hiring engineers or sales? We'd love to show you how it works. Benchmark details: https://lnkd.in/gt_JEKDE Learn more: https://www.perfectly.so

18

3 Comments

 Been working with Claude Code sub-agents. ceated somethomg usable - 4 production-ready configuration files that you can copy-paste today to transform your API development workflow. The specialized agent approach is delivering some impressive results: 6x faster development with 85% fewer errors.  Instead of asking one general AI to handle everything, you get focused experts for each domain:  • API Design Specialist - handles OpenAPI specs and mock generation  • Contract Testing Specialist - catches breaking changes before they hit production• Security Specialist - runs OWASP validation automatically  • Performance Specialist - detects regressions in real-time  What used to take 185 minutes per API change now takes 15 minutes. The math is compelling: specialized_expertise × automated_workflows = exponential_quality_improvement.  These aren't just examples - they're production-tested YAML configs you can drop into your .claude/ directory and start using immediately. The post includes complete configurations with detailed prompts, tool access, and quality gates.  The results speak for themselves: real-time documentation sync, automated quality validation, and significantly reduced cognitive overhead.  What's your experience with specialized AI workflows? Have you tried domain-specific approaches in your development process?  https://lnkd.in/d2enES7G

7

For understanding the virtual thread state transitions... Specifically, thread pinning on Java 21 vs. Java 24 and when the thread waits on IO how the platform thread views the state compared against the virtual thread enabling interruption

3

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

Rolled out Jan 1. California residents can now continuously opt-out of 500+ data brokers for free. https://lnkd.in/ephsgD-K Happy new year from the California Delete Act

8

4 Comments

We shipped automatic snapshot expiration for R2 Data Catalog today. With this new feature you can define retention policies (age thresholds and minimum snapshots to keep) so old snapshots are cleaned up automatically, reducing metadata overhead and storage costs while keeping table performance healthy. https://lnkd.in/eAD5rKd8

46

With AI, according to all of the marketing materials from all of the companies, including Meta, SWE not required anymore and you can just scale your engineering up by using LLM models alone... Yet: - Meta cannot keep Desktop app for their own Facebook Messenger alive - https://lnkd.in/ekKTbEfG keeps growing and Google cannot continue to develop already released products - Windsurf cannot keep up and to have their IJ plugin for Remote Development in sync with their non-RD plugin https://lnkd.in/eeMjrZ_6 - Claude IJ plugin is also far from perfect What has happened to "You need to practice what you preach"?

15

1 Comment

The Claude ecosystem is exploding faster than I expected. My newest favorite plugins: - Get Shit Done (orchestration layer that plans and builds apps phase by phase) - Ralph Wiggum Plugin ( iterative, self-referential AI development loops) - n8n-MCP (direct Claude integration with 1,000+ automation nodes) What's interesting is how quickly we've moved from simple AI coding assistants to full orchestration platforms. The shift feels similar to what happened with CI/CD tools a few years ago. First you had basic scripts, then comprehensive pipelines, now full DevOps orchestration. We're seeing the same evolution with AI coding tools, just compressed into months instead of years.  #AI #Claude #SoftwareDevelopment #AIAgents #DevTools #Automation

10

There is a lot of talk about AI replacing developers (with many thousands of layoffs from large companies seemingly lending credence), but what if there's a twist? These two stories suggest a potential case of incoming dramatic irony: the very power consumption plus business model that makes AI coding tools so useful and intriguing might also make them unsustainable at scale: 1️⃣ There will be throttling and pricing tiers, possibly leaving business in the lurch as they hit those limits: 🔗 https://lnkd.in/gEAQfapb 2️⃣ Meanwhile, the infrastructure costs behind AI tools are real and rising. This isn't a free lunch; it's a business product, again suggesting the inevitability of pricing tiers and throttling: 🔗 https://lnkd.in/gzaJHbYy What if the eventual "plateau of productivity" for AI coding includes the realization that the best ROI on development still means a team of skilled, intelligent, reliable, engaged humans?

1

1 Comment

I often see comments along the lines of "I can't believe anyone still uses ____ tool for development" or "anyone that thinks that clearly doesn't know what this thing means" I just see this as incendiary and demoralizing. it is never seeks to understand a position, but dismiss it as an objectively a bad position. The development world is rife with imposter syndrome and other mental issues. Comments like these can easily make someone second guess their own competency. Let's strive for understanding, and suggest alternatives compationately, not confrontationally.

6

I wrote an article on how we translate user-generated content at scale at Mercari. It summarizes two years of work with various models, LLMs but not only, and how cost evolved, reducing it 100x over the period. I also go over non-AI features and user experience. https://lnkd.in/gbpQAQmX

39

2 Comments

How to disrupt an industry with AI? On the latest Hard Calls, Trisha Price sits down with Vantaca CEO Ben Currin to explore how they’re reshaping community management with AI-powered workflows.

5

Developing a large ecosystem with both open- and closed-source components comes with some interesting challenges. Back when we were bootstrapping CKEditor, my purist mindset pushed us to develop each of our 30+ packages in separate repositories. The goal was to sort of "force" higher modularity and more responsible architecture for every component. As you can probably guess 😅, it did not scale. Later, we consolidated most of the packages into two monorepos – one open source, one closed source – while keeping a few granular packages separate where it still made sense. This helped a lot, but it didn’t fully solve the problem. We were still working across multiple repositories and kept running into issues like dependency coordination and lockfiles. Recently, we took another step forward. We migrated to pnpm and redesigned our overall approach. In this post, Filip Sobol explains the new setup implemented by the CKEditor 5 Platform team and what we learned along the way: https://lnkd.in/eHS28qxe

22

3 Comments

Measuring developer productivity is complex. At Meta, the team has been taking a rigorous, data-driven approach and even building new infrastructure to support it. Kudos to Payam, Pavel, and the entire team for this impressive work and presentation. Over the past year, my main focus has been on building AI coding agents, so this really resonates. Check out the section about Devmate at 6:52 in the video to learn more on what I've been up to. A few highlights: - Thanks to Devmate, Meta’s developers are more productive - with Diffs per Developer per Month (DDM) improving by 6–12% (a conservative estimate!). - Devmate tracks AI code attribution at the character level, allowing Meta to know precisely which code was generated by Devmate vs. written by humans.

15

Google Changed RAG with NEW Gemini File Search Tool (RAG Made Simple) 5 Days Ago, Google Took a Big Step Forward in the RAG Ecosystem Google has launched the File Search Tool, a fully managed Retrieval-Augmented Generation (RAG) system built directly into the Gemini API. This new capability abstracts away the complexity of retrieval pipelines, offering a simple, scalable, and developer-friendly way to ground Gemini models with your own data. Key Highlights: 1. Fully managed RAG system: No need to build or maintain custom retrievers or vector stores. 2. Integrated & scalable: Works seamlessly within the Gemini API. 3. Improved grounding: Delivers responses that are more accurate, relevant, and verifiable. 4. Cost-efficient billing: Embedding generation at query time is free; you only pay for initial file indexing at $0.15 per 1M tokens. This launch marks a major leap toward making enterprise-grade RAG systems accessible to all developers, empowering teams to build intelligent, data-grounded applications faster with precision and transparency. #GenAI #GoogleAI #Gemini #RAG #AIInnovation #LLMs #AIEngineering #GeminiFileSearchTool

7

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

Kagi has a translator that will turn English into McKinsey-speak. Also Gen Z. The avian asset’s presence facilitated a positive shift in emotional sentiment, driven by its high-gravity aesthetic and disciplined posture. I initiated a stakeholder inquiry: "Despite your suboptimal grooming and lack of plumage, you demonstrate significant risk appetite. As a legacy entity from the nocturnal ecosystem, what is your core brand identity on the Plutonian shore?" The Raven responded with a definitive, non-negotiable strategic pivot: "Nevermore." https://lnkd.in/ezKWuRxQ

1