I read some stuff by Zitron a while back and even back then it seemed untenable and divorced from reality aiming mostly to score points with an audience that’s more than eager to believe him.
Knowing what to build, knowing when the system is telling you something, knowing when to cut, knowing when the fast path is the wrong path — these were never just “execution.” They were always the harder parts and to pretend they’re solved by AI is questionable, to say the least.
This is an interesting realignment of the org chart. I’m not sure whether we’ll reach that endpoint, but it’s obvious to divide the team into the factors laid out in the article.
One person helping another because they have a relationship and the work needs doing. Sean frames these as workarounds. Necessary but awkward and illegible. But what if they’re not workarounds? What if they’re the actual thing?
A quick little interview with the creator of Tantivy a Rust based search technology that I’m hearing incresaingly more about and which I’m very eager to try out in production.
After hearing so many good things anout it, we normal people finally get to use a SAP solution as well. I wonder how quickly they’ll get out a first version for us to test and how wonderfully usable it’s going to be.
Here’s an idea: Developers should go through a proficiency test before they are allowed to use LLMs to generate code. Think of it as an AI driver’s license.
“This amazing snippet of code was languishing in the docs of dead software, which in turn the original formula was scrawled away in a math textbook from the 60s.”
To me, an obvious benefit of using jj is the power of the tooling and how intuitive it is to use. Once you have internalised the mental model, it doesn’t take many commands to achieve what you want.
Using Postgres logical replication to stream a video call through the database is a brilliant hack that displays both the power of the database and of standard web applications.
People who need them being able to vibe out essential accessibility apps using just Claude Code is definitely making a difference. It’s hard to argue against this.
If you removed hand-written code but didn’t add explicit invariants, you lost rigor. If you’re generating implementations without rigorous evaluation, you lost rigor. If you’re accepting code because it runs rather than because you understand it, you lost rigor.
More stuf by Sean Goedecke that I came up with independently myself. Being unreasonably effective is one of the best ways for career success. Getting that way requires critical thinking and consistent application.
how Amazon’s traditional viciously customer-focused approach to development was a poor match for the research-style work that was required for developing AI
No individual in the organization will ever have an accurate mental model about how the entire system works. To put it simply: It’s the stuff we don’t know that bites us. There’s always stuff we don’t know.
I think I have to disagree with this post and call it mostly wishful thinking. I haven’t seen that much adoption of vibecoding in Rust. Most of the data points I’m looking at point towards Go being the clear leader here with its low expressiveness, quick compilation and rather basic type system.
The goal is not to embed Django inside another system, but to demonstrate that Django ORM can already act as a lightweight, standalone data access layer.
It’s obvious that international platforms will need to support international writing. The world is a big place and ASCII hasn’t been enough for a while now.
It sounds so simple but the amount of work and trade-offs involved in getting to a single list of priorities is where a lot of the difficult work lives.
We’ll have so many of these AI exploits coming out that there are probably exploit groups specializing on them. It’ll be a very dependable stream of revenue for them.
Picking between Rust and Python, the code is by now not even that relevant anymore but stuff like easy to distribute and instant startup time is very hard to argue against.
Automatic binary analysis using AI is not nearly good enough yet to be usable but it is improving and increasing the accessibility of doing this kind of analysis.
Then he said, “You know what the difference is between you and me? I know I’m a mercenary. You thought you were an artist. We’re both guys who type for money.”
if your best idea for what AI can do in the workspace is “replace a hundred human beings with a server rack doing the same thing”, you’ve got no business calling yourself a techno-optimist
I kept hearing so many good things about Go that I decided to make the jump from Rust. I had Claude port my app1 and couldn’t be happier with the result.
Somebody has gone through the effort of creating a Github uptime page where the company itself can’t be bothered to. It’s bad but downtime is mostly concentrates in Actions and Copilot.
Microsoft, the company known for such amazing achievements as Teams, Github’s uptime, Copilot etc. has managed to add features to Notepad in such a way, they introduced a remote code execution vulnerability.
There was an effort to update the 12factor principles but reading this you can see how well the factors have held up over the 15 years they’ve been out there.
I really don’t know why I write Go (a bit too much of a purist) because as described here it really is quite good, enjoyable and extremely LLM compatible.
A software library that has no code but is generated ad hoc by LLM is such a brilliant provocations and an end run to the natural conclusion of these developments.
Europe has lots of significant industries and capabilities. There is nothing stopping us from having a competitive local cloud offering (except of course the lack of political will).