August 12, 2025

Tuesday Letter #37 - Focusing on Automation

Hey, Happy Tuesday!

Why are you getting this: You signed up to receive this newsletter on my personal website. I promised to send you the most interesting sites and resources I have encountered during the week. If you don't want to receive this newsletter, feel free to unsubscribe anytime.

Personal Updates

• This week's issue is a little heavy on the content. I'm experimenting with how much I share in these emails. If you have any thoughts, please reply to this email. Would love to hear some feedback.

• Focusing on automation these days. In last update Ask HN Digests was fully automated. Now trying to automate Built with Django newsletter. Almost done. My subs deserve consistent emails.

• After a nice day on the farm with my 2 year old son, I accidentally dislocated his arm. Spent the rest of the day in the emergency room. The line was huge, thankfully some nice people let us through and we fixed it in a second.

• Started using Wispr Flow last week. Getting more comfortable dictating instead of typing. Definitely recommend trying it out. Super useful when coding with AI.

• Made good progress on automating Marketing Agents. Also, decided to pivot and make it just about content not a whole marketing suite. Hopefully that will help me ship a final product much sooner. And more importantly focus on the main feature and then marketing the product to get some users.

• Finished reading "Growth Levers and How to Find Them" by Matt Lerner. Definitely a good read for founders who found their PMF and are looking to grow. Probably a little to early for me. Also, shared the summary on my site.

• Published my notes and learning for Start Small, Stay Small by Rob Walling

Quote of the week

"I've got this working theory that you can tell how enthusiastic someone is about a subject by how obscure their favorite book is. It's just a simple numbers game: there are more obscure books than popular books, so there are more obscure good books than popular good books, so the more books you read the more likely your favorite is an obscure one."

— Computer Things, Why You Should Read "Data and Reality"

This hit me hard because it's such a clever way to think about genuine expertise versus surface-level interest. I've noticed this pattern in my own life – the people who really know their stuff in any field rarely cite the bestsellers as their favorites. They've gone deep enough to discover the hidden gems that casual enthusiasts never encounter. It's not about being contrarian; it's about having put in the reps.

Cool person I encountered this week

Onur Solmaz

I discovered Onur through his "Typed languages are better suited for vibecoding".

Onur is a self-taught programmer and Founding Engineer at TextCortex. Beyond his day job, he's created some fascinating projects including a simplified German dialect called Alman andd contributes to Manim Community. You can find him on Twitter, GitHub, and LinkedIn.

Books I read this week

• Growth Levers and How to Find Them: 28610 words (Finished)
• Alexander Pushkin Anthology: 19623 words
• Full-Stack Tao: 11999 words
• The 22 Immutable Laws of Marketing: Exposed and Explained by the World's Two: 4897 words

Articles I read this week

• Why You Should Read "Data and Reality" by Computer Things - "Data and Reality" by Bill Kent explores how we interpret and represent information about the world, not the world itself. The book challenges readers to think deeply about identity, meaning, and the complexities of modeling reality. Although out of print, the original edition is available online and highly recommended for anyone interested in data and information systems.

• Software books I wish I could read by Computer Things - The writer loves how deep books can be and wishes more existed for software work. They list dream titles on things like configuration, data schemas, CS basics for engineers, MISU patterns, key tools, old optimizations, and Sphinx internals. If someone writes these books, the writer promises to buy them.

• The skill $10k/month writers keep secret by Write With AI - Six-figure writers earn more because they master sales, not just writing.
  Good sales follows five checkpoints: open, discover problems, teach, pitch, and answer objections.
  Recording calls and using the AI prompt helps writers review and practice to close more deals.

• How I like to write with AI by Nathan Barry - Nathan Barry dictates his ideas and runs the transcript through Google Docs, Claude, and NotebookLM. Claude trims, clarifies, and adds counterpoints, while NotebookLM supplies real examples, making the whole process faster and sharper than typing. He urges creators to try this talk-first workflow and adds quick plugs for a LinkedIn growth podcast, a Linktree-Kit app, and Ed Latimore’s new book.

• Claude Code by Grant Slatton - Claude-code is a tiny command-line wrapper that runs Claude in a loop and reads rules or data straight from your files.
  Storing key instructions in CLAUDE.md and adding simple workflows (for context gathering or calling other models) greatly improves its accuracy and reach.
  Because it’s clear and hackable, users automate steps, spin up parallel or nested “sub-claudes,” and keep tweaking the setup as needs grow.

• LLM Memory by Grant Slatton - AI models lack true memory, so keeping facts tied to time and space is tricky.
  The author favors document-based knowledge graphs where episodes and meta-documents connect and summarize information for later retrieval.
  Layered storage, intelligent traversal, and selective forgetting may mimic human memory as context windows expand.

• Day 47 of Claude Code god mode by Onur Solmaz - Since adopting Claude Code in May 2025, the author now speaks ideas and sees working code almost instantly. In 47 days they merged repos, built new tools and apps, and finished projects that once took months. They say this “agentic” AI workflow will soon reshape software work and reveal who can’t keep up.

• Auto-generating pull request documentation with Claude Code and GitHub Actions by Onur Solmaz - Anthropic’s new GitHub Action lets Claude Code automatically write documentation after each pull request is merged. You add a simple workflow and your API key, and it creates a follow-up PR that stores the docs in the repo. This makes changes easy to understand for both people and future AI tools, although the service still costs quite a bit.

• It’s the Housing, Stupid by Nick Maggiulli - Sky-high home prices and mortgage rates have pushed even wealthy would-be buyers to keep renting and stash their cash elsewhere. That sidelined money is flooding both money-market funds for safety and meme stocks for thrills, causing today’s strange mix of record cash balances and 2021-style speculation. The pattern may linger until housing either gets cheaper, rates drop, or much more housing is built.

• I Shipped a macOS App Built Entirely by Claude Code by indragie.com - The author built a macOS app almost entirely using Claude Code, an AI coding tool that writes and tests code quickly. Claude Code works well with SwiftUI and can improve with clear instructions and detailed feature specs. This shows a future where AI-driven coding tools could change how software is made.

• 6 Weeks of Claude Code by Puzzmo Blog - Claude Code has greatly improved how I write and manage code by quickly creating large sections instead of line-by-line work. In six weeks, I completed many complex projects alone while still handling my usual tasks. This tool acts like a patient programming partner, speeding up development and easing the work.

• Agentic Coding Things That Didn’t Work by Armin Ronacher - The author tried many agentic coding tools but found most automations didn’t stick or improve his work. He prefers simple speech-to-text interaction and manual context sharing over complex commands or hooks. Automation must be tested carefully, or it risks making users less engaged and lowering code quality.

• How to get recommended by ChatGPT by Kyle Poyar - AI search tools like ChatGPT are changing how buyers find products by giving direct answers instead of links. To get recommended, brands must create very specific content, get mentioned on trusted sites like Reddit, and use structured lists and tables. Measuring success means tracking how often your product is talked about, not just clicks, and staying updated as AI changes.

• You should be playing with AI agents for marketing by Kyle Poyar - AI agents can handle many marketing tasks, working like a team of helpers that boost productivity. Jacob Bank runs a marketing org with over 40 AI agents doing the work of five people. Anyone can build and use AI agents to automate tasks and improve marketing results without being an expert.

• o3 used my saved Pocket (RIP) links to profile me by noperator.dev - The author used an AI tool called o3 to analyze nearly 900 saved Pocket articles and create a detailed personal profile. The profile accurately inferred their age, location, career, family life, and interests from just a list of URLs. This shows how powerful and accessible data analysis has become for personal insights and recommendations.

• The Force-Feeding of AI on an Unwilling Public by Ted Gioia - Tech companies are forcing AI into products without asking users, even raising prices for unwanted features. Most people don’t want AI and only a few would pay for it voluntarily. This forced use risks losing customer trust and removes real human choices in everyday life.

• Whisperglass by Justin Duke - Justin Duke finally started coding Whisperglass, a small Django-based SaaS meant more for learning and sharing than for profit. He built the core features in a few weekend hours using AI help and code from Buttondown. He plans to keep it open source and welcomes readers or testers to follow along.

• Using Claude in Github: Hype and Reality by Cory Zue - Claude can now be easily linked to a Github repository using Anthropic’s Github Actions to get near instant coding help. Writing clear instructions for the AI is hard, which limits its usefulness on vague tasks. Even though it jumpstarts projects quickly, its code often needs manual tweaking to meet quality standards.

• Digging into Ducklake by rmoff's random ramblings - DuckLake is a new specification for storing data as Parquet files with a SQL database for metadata. It bundles storage and catalog management to simplify data engineering tasks. It is experimental and currently available only as a DuckDB extension.

• Practical AI techniques for daily engineering work by seangoedecke.com RSS feed - Engineers can use AI to get a quick second opinion on their code and spot simpler solutions. AI helps create short debugging scripts that save time during problem solving. It can also gather evidence or fill technical gaps when needed.

• When o3 Plans Your Career Better Than You Do by Every - The article shows how an AI tool, o3, can design a detailed career roadmap. The AI suggests bold, creative paths that feel authentic. It also offers steps and prompts for using AI as a personal career coach.

• AI Brainrot means developer opportunity by Doug Turnbull - AI tools let you build apps quickly but can leave you with a half-understood product. This speed raises expectations so that even small glitches feel like huge problems. Savvy specialists can profit by fixing these inconveniences that others now find unbearable.

• Why Generalists Own the Future by Every - Generalists can quickly learn and adapt in the fast-changing AI age. They solve new, unclear problems using a mix of skills from different fields. This diverse approach helps them ask the right questions while AI handles routine tasks.

• What is OpenTelemetry and how to add it to your Django application by Jessica Garson - OpenTelemetry is an open-source framework that helps you monitor applications in a standardized way. This guide shows how to add OpenTelemetry to a Django to-do list app, connecting it to an Elastic backend. You can choose between automatic and manual instrumentation to customize how you collect monitoring data.

• Why it's nice to compete against a large, profitable company by Jason Cohen - Large companies rely on profitable business lines that lock them into high prices and slow change. Small startups can win by using lower costs, new tech, and fresh ideas to target those profitable segments. However, attacking a loss-leader won’t work because the big firm can afford to sacrifice losses in that area.

• The Cline AI Assistant is Mesmerizing by mtlynch.io - The author tried the Cline AI assistant and was mesmerized by how it fixed code automatically. It’s thrilling that AI can now write and revise software so proficiently, yet also unsettling for developers’ futures. The author spent hours refining a tool with Cline and shared practical insights into using AI effectively.

• My 25-year adventure in AI and ML by Austin Z. Henley - Austin Z. Henley reflects on his 25-year journey in AI and machine learning, which began unexpectedly while developing games and software. He shares milestones from his programming projects through his academic career and research at Microsoft, highlighting the evolution of his skills and interests. Now, he continues to teach and create AI tools, looking forward to future innovations.

• Preferring throwaway code over design docs by Doug Turnbull - Doug Turnbull argues that coding should often take precedence over designing detailed documentation. He suggests using draft pull requests for early feedback and rapid prototyping, emphasizing that real insights come from coding rather than static design docs. Ultimately, he believes that a disciplined coding approach can lead to more efficient development than traditional design methods.

• Revisiting uv by Kevin Renskers - Recent updates to uv have resolved major issues that previously made it less appealing than Poetry, such as supporting dependency groups and showing outdated packages. Users can now manage dependencies more effectively and see which packages need updates. Despite some minor drawbacks, the improvements in uv make it a strong choice for managing Python projects.

Other cool links

• Podscan Ideas: The Best Business Opportunities from Millions of Podcasts
• Historical Tech Tree: Interactive visualization of technological history
• 2025 How to Grow Almost Anything | Notion: HTGAA '25 - MIT/Harvard Synthetic Biology online course
• KathaaVerse: KathaaVerse: Transform Any Book Into your own Interactive Adventure
• Answer.AI - Practical AI R&D – Answer.AI: Practical AI R&D
• WallStreetBets Top Trending Stocks: A look at which of Reddit's WallStreetBets stocks are trending and view historical mentions for any ticker.
• Free Ideas — Quarter Mile: freeideas.net There’s a quote I like from architect Christopher Alexander: “Really one of the very largest problems that is facing the Earth just now...

Support

You can support this project by using one of the affiliate links below. These are always going to be projects I use and love! No "Bluehost" crap here!

• Buttondown - Email newsletter tool I use to send you this newsletter.
• Readwise - Best reading software company out there. I you want to up your e-reading game, this is definitely for you! It also so happens that I work for Readwise. Best company out there!
• Hetzner - IMHO the best place to buy a VPS or a server for your projects. I'll be doing a tutorial on how to use this in the future.
• SaaS Pegasus is one of the best (if not the best) ways to quickstart your Django Project. If you have a business idea but don't want to set up all the boring stuff (Auth, Payments, Workers, etc.) this is for you!

Sponsors

This newsletter is sponsored by Marketing Agents. Well, sponsor is a strong word. It is just another project of mine that I wanted to share with you 🙈.

If you have a side project and are struggling with the marketing side, it might help! It's still in early dev, so any feedback is super useful.

If you want to become a real sponsor, just reply to this email 😄