the padded cell

#björn

15 August, 2025

DIKW pyramid when working with genies

…classification of different types of understanding

Curtis shared the DIKW pyramid, that I hadn’t heard of, and said that he feels that the genies helps him process the first two layers much more quickly than he can alone, that it can somewhat help penetrate into knowledge but not much. Which allows him to spend less time in D/I (collection) and then more to then K/W (processing?)

The pyramid of knowledge with a base of data, the next layer built of information, then knowledge and finally wisdom at the peak.

[… more]

If you start a non-interactive bash shell it will source the content of the file defined in BASH_ENV (and ENV for a POSIX shell).

When bash is started non-interactively, to run a shell script, for example, it looks for the variable BASH_ENV in the environment, expands its value if it appears there, and uses the expanded value as the name of a file to read and execute. Bash behaves as if the following command were executed:

if [ -n "$BASH_ENV" ]; then . "$BASH_ENV"; fi

but the value of the PATH variable is not used to search for the file name.

YAML does not support unsplatting lists (basically, merging list items inline like you can with objects) and that’s intentional.

So if you have a document like below, there is no syntax to make the commands a three item list:

---
example: &example
  - "Hello"
  - "World!"

name: "hello"
commands:
  - << *example
  - "Oho!"

WILL NOT turn into:

---
example: &example
  - "Hello"
  - "World!"

name: "hello"
commands:
  - "Hello"
  - "World!"
  - "Oho!"

…and just because before I found the issue where it was described that this isn’t happening I had made a test repo to try and understand if it was a library/usage issue, because I noticed I could get things running on CI with << *example which syntax didn’t give syntax error, but also did nothing when running on Drone CI.

13 August, 2025

/usr/bin/env executes commands with flags/subcommands, not just bare executables. Which is great if you, for example, have a script/lint that’s a Python script, and it needs dependencies from a virtualenv that isn’t active when you call it.

Just put your shebang as /usr/bin/env uv run python3 and it always runs in the virtualenv, no wrapper script needed. This feels obvious in hindsight, it’s what you expect from these tools 😃

12 August, 2025

Anthropic on prompting for agents

…and musing on genies possibly teaching empathy

Anthropic has released some recordings from their Code w/ Claude event in May and the Prompting for Agents presentation’s “key principles” are basically to empathise with your agent, imagine it’s a brilliant new grad, book smart but missing all things practical at their first job: You need clear concepts, unambiguous instructions, and well-named and designed tools.

I’m more than a little amused that we’re basically coming to a point where, if things pan out with the genies, then the best wranglers will be the ones that can empathise the most with others. Then again, I always thought the brilliant asshole was the exception, they only survive if they’re in charge or are legacy to the company. And sitting in on the review meetings at work… it’s definitely the ones that help the team that we like the most, even if we sometimes need to push them to have their name on something so people don’t overlook them.

[… more]

10 August, 2025

When everyone gets a genie

learning to handle expertise you didn't earn

We’ve always had to figure out who to trust for expertise, but until now, that access wasn’t universal. Used to be the bloke at the pub who ‘knew things,’ or that friend of a friend who could fix computers (me as a kid), or books if you had the patience. Rich people had their ‘real’ experts on call, though Bernie Madoff shows how well that could work out.

The internet changed things, sure. Suddenly you could reach out to actual experts, find communities, and learn from people across the world. But you still needed judgment, because you could also find a community that agreed the earth was flat, so, you know, mixed bag.

What hasn’t changed is we’ve always had to figure out who to trust. What’s changing is how we get our answers, and who gets to sound like an expert.

[… more]
blog 10 min read #thinking-out-loud, #ai, #genie, #learning

02 August, 2025

Drone CI has special handling for ${VAR} and will replace it itself instead of letting the shell do it. So if you have export PATH=${GOPATH}/bin:$PATH and have $GOPATH defined in your global environment key, then it’ll become a blank value. But if you do export PATH=$GOPATH/bin:$PATH it’ll work. You can escape by adding an extra dollar sign, i.e. $${VAR} will be evaluated by the shell.

This is different from normal shell where ${VAR} is the same as $VAR and it exists to make sure you can concatenate strings without issues, i.e.

VAR=m000
echo "'${VAR}est'"  # => 'm000est'
echo "'$VARest'"    # => ''
til Updated #drone-ci

Drone CI will not smartly skip subsequent steps if you have a when on the very first step in a depends_on chain, so you have to repeat the when condition for each step because it doesn’t realize that the first dependency is gone.

Drone CI’s when for deciding in which cases to run steps/pipelines targets the merge target branch for PRs and not the actual PR’s, also it pulls from refs/pull/<num>/head instead of refs/heads/<branch> so you can’t target the branch itself using the pull_request event (use push on the pipeline and then branch for the step instead).

In shell scripts if you do "$@" it will actually expand “quoted sentences” correctly, and if you just do $@ it will always unwrap them into single words, I thought that if you did "$@" it would combine all arguments into a single argument, and what it does is do what I thought $@ alone did.

I.e., with "$@" the arguments "one two" three will be 2 arguments, the first being "one two", and without it will become three arguments, all separated by space.

The <link rel="canonical" href="…"> is a bit like a correlation ID between services and I need also to have it on my canonical page, because if someone sends it with ?utm_source=foo it could be a different page than the one without the query string.

The scripts/commands inside a shell script inherit access to STDIN when you call them, so if you have a shells script that only has cat and you do ./script.sh < script.sh then it’ll output the content of itself

You can view the webhooks that GitHub sends in your project settings, so for example Drone CI getting notified and exactly what information that goes into the payload, and when/if it was sent, is available there.

Go to Settings -> Webhooks and then click on the integration you want to look at.

01 August, 2025

You can enable the fingerprint reader for sudo on macOS, and pressing my finger on a button beats having to type the password, steps:

  1. cp /etc/pam.d/sudo_local{.template,}
  2. Edit /etc/pam.d/sudo_local and uncomment auth sufficient pam_tid.so.

The reason for doing this in sudo_local is that this file will not get reset with system changes from Apple.

24 July, 2025

Your name is still on it

learning to ride the AI motorcycle without crashing

A colleague recently said something that’s been rattling around in my head: “AI gives you speed, but it doesn’t give you direction.” And the more I use these tools, the more I think that undersells the danger.

I have been wondering how to think about AIs (or genies) and how computers are like bicycles for the mind, as Steve Jobs put it, and I think these tools take it further. They are more like motorcycles for the mind. They go really fast, and you better not treat them like a bike, because you need to know what you’re doing. How to handle that thing. You need to make sure you don’t try to go too fast too soon, or for too long, because you’ll get speed blind and… things will happen.

[… more]

12 July, 2025

Which hat are you wearing?

...you wouldn't wear a beanie to the beach

I was in an incident review recently where one of the problems was a human going too fast. This process is very manual, repetitive, and boring, and it rarely fails, so we skip some steps. That mostly works (see Why Do Things Go right?), except when it doesn’t. Ripe for occasional issues and likely in need of automation.

I believe we often skip steps because we don’t know why we do them. It’s not tedious for tediousness sake, it’s often there because it’s important. And until we have the automation, or maybe we’ve intentionally chosen not to automate it, we need to find some way of helping the human ’live the situation.'

[… more]