“The most important skill in prompting is expertise in the domain you’re prompting for.” I have nothing to add except yes, amen, and read this post: www.seangoedecke.com/llms-reward-...
LLMs reward expertise
seangoedecke.com
Christian Minich
@christianminich.com
Dad, data, process, philosophy | Forward Deployed Engineer @Vercel | My wife says my headstone will read “sorry I’m late”
I’ve seen a lot of skepticism about AI similar to the famous “you can see the computer revolution everywhere but in the productivity statistics.” Everyone is being told to adopt AI at work, but to what end? Where are the outcomes? This paper by Ethan Mollick has an answer:
I’ve been on team “skills are all you need” for a minute, but stateless MCP + recent model advances are changing that, much like @simonwillison.net explains in his blog: simonwillison.net/2026/Jul/31/...
Stateless MCP has recaptured my interest (and inspired mcp-explorer and datasette-mcp)
Tuesday was Stateless MCP day—the rollout of MCP 2.0, or the 2026-07-28 Model Context Protocol specification to use the more formal but less memorable name. This is the most significant …
simonwillison.net
Agents forcing SaaS to be hackable by humans too is my favorite side effect of AI app.notion.com/p/notionamba...
How to use Notion as Code | Notion
Notion as Code allows you to build and update Notion programmatically in bulk. Instead of having to make individual public API requests, you can describe the final state and we handle updating your wo...
app.notion.com
It’s only a company brain if it comes from the SF region of America. Otherwise it’s just a sparkling data warehouse
On the other hand, in my experience most “migrations” don’t look like this and humans will be critical for a very long time. Let’s say you are migrating your app codebase from Python to Go. The agent swarm (take Cursor’s example) got 80% of the way there.
Christian Minich@christianminich.com · 2w agoI think some people are missing the point of these “my agentic harness migrated this library to rust” posts. These are interesting benchmarks precisely because they are a known, difficult, solvable problem that has objective grading criteria (i.e. the test suite and “it compiles and runs!)
I think some people are missing the point of these “my agentic harness migrated this library to rust” posts. These are interesting benchmarks precisely because they are a known, difficult, solvable problem that has objective grading criteria (i.e. the test suite and “it compiles and runs!)
Excited to share that this week I joined Vercel as a Forward Deployed Engineer! Vercel is building the future of agentic infrastructure, and I can’t wait to work with some of the largest organizations in the world as they transform their organizations with AI.
There are like only three design patterns you need to use 80% to make stuff good and fast and elegant Make it a lookup (hashmap) Split it up into smaller pieces (sharding) Combine it into one large piece (map reduce)
It was great to join @christianminich.com in this webinar. He shared a lot of great insights in how he's using AI for data engineering.
dagster.io@dagster.io · 4w agoIf you missed today's webinar where Christian Minich dives deep on how he leverages AI when building projects for prospective Dagster users, then you can find the recording here! www.youtube.com/watch?v=_o7d...
Is there any trivial thing more annoying than waiting all day for a FedEx package you have to sign for?
The real AI transformation was the bog standard process improvement methodology actually adopted by teams along the way.
Teaching my kids about space travel the best way possible: listening to Space Oddity on repeat
This Thursday I’m going to make a Dagster project from scratch, LIVE. This Thursday, I’ll show you, LIVE, how we use AI at Dagster draw the rest of the owl. Interested? Sign up here: app.sequel.io/event/d64257...
Mr Andersen, your LinkedIn posts have recently fallen below the standard 100% positive sentiment score to 98.3%. We believe you have bypassed your LLM, and are writing posts by hand. Please reset your LLM to the standard configuration to continue to enjoy the benefits of existence. Have a great day!
y'all read anything insightful about how we change software development methodologies in light of really useful LLMs? it feels like the bottleneck moved from "pounding out code" to "figuring out what problem to solve". I think devs should be at the keyboard less and sit with users more.
This by @mmullins.coginiti.co is so good. Summary: philosophers have had a place, especially around metaphysics, in tech, but with LLMs epistemology and ethics take center stage and are no long just a necessary foundation, but a product differentiator open.substack.com/pub/msmullin... #dataBS
Philosophers in the Machine
Philosophers are not new to AI. From ontology engineering to alignment, philosophy has long shaped how machines classify, reason, and act.
open.substack.com
Data engineers to software engineers who are all excited about context pipelines. (AI engineering is a data engineering shaped problem) #dataBA
This sentence could have been a 4000 word essay with perfect pacing and a thousand honstly's
If you think AI generated posts on LinkedIn are bad, wait till you see AI generated spreadsheets with all hard coded values
I’ve worked hybrid since 2017 and then fully remote since 2020. Seeing my kids to school virtually every day, being able to take them to appointments and school events, and just being present is something I wouldn’t trade for the world.
My wife when she “forgot” that she has 5 shrub stumps to dig out today instead of 3
cloud.google.com/blog/product... I sometimes like to imagine that AI is a psyop just to get people to improve documentation by elevating markdowns to “serious engineering”
How the Open Knowledge Format can improve data sharing | Google Cloud Blog
Learn how the Open Knowledge Format helps secure data sharing and improves collaboration across teams with standardized documentation.
cloud.google.com
A whole Vergecast about Markdown? Inject this into my veins www.theverge.com/podcast/9500...
Today’s Vergecast: # The **epic** story of Markdown.
Markdown is a system for writing that makes it readable to both humans and computers. It’s all about the symbols. You use - to make a list, * for emphasis, ** for even more emphasis. Brackets and pare...
theverge.com
When I ran a team of analytics engineers, I told everyone I interviewed that we followed the Ratatouille style of building a data warehouse: instead of “anyone can cook” I said “anyone can contribute.” In the film, that phrase is widely misunderstood to mean “everyone can cook” by other chefs.
I've started putting a "Human TL;DR" at the top of AI generated PR descriptions. The AI descriptions are useful and re-writing them feels like a waste of time, but the human TL;DR means the reviewer can scim the AI part for the parts they find of interest.