It's the First day of my batch today. We <del>haven't had the kickoff meeting yet</del> but
I'm taking some time to set intentions for the week.
This week is inevitably dedicated to figuring things out, and I don't think I
should expect to get a lot of programming done, but I do still want to set some
goals.
<ul>
<li>
Figure out the shape of my days: I'm 6 hours ahead NY time, so core hours fall
between 5-11pm. Can I get some work done in the morning? How can I ensure I'm
productive in the evening? How do other remote recurses do it?
</li>
<li>
Figure out the shape of the full batch. I have some project ideas, but I'm
going back and forth between depth and breadth. I'm hoping that after intro chats
and meeting other recurses this week I'll have a better idea of what's possible
/ realistic.
</li>
<li>
I would still like to get some small project done. A big explicit goal for me is
to build shipping muscle, meaning once I set out to do something, finish it,
write about it and present it. I would like to have something small and
presentable by the end of the week.
</li>
</ul>
A couple of meta concerns I have:
<ul>
<li>
How can i make sure I get a healthy amount of movement in? Given the timezone
difference, I won't get a lot of time to skate/surf/bmx in the next 6 weeks.
I've grabbed a jump rope, yoga mat and some resistance bands to make sure I get
some workouts during the day.
</li>
<li>
What's a good ratio of blogging to coding? I want to use blogging as an
accountability method, but I know how easy it is to spend way too long on a
piece of writing...
</li>
</ul>
On that note, piece of writing done. Time to get to work.

It's the First day of my batch today. We haven't had the kickoff meeting yet butI'm taking some time to set intentions for the week.This week is inevitably dedicated to figuring things out, and I do...

Recurse week 1 Intentions

I just got confirmed for the Summer 2 batch at the <a href="https://www.recurse.com/">Recurse Center</a> starting in
July 1st and I'm through the moon.
I have been wanting to attend for the longest time (I have tweets from as far
back as 2017) but always had other commitments. Back in May I decided it was
the perfect time, got my application together and after a couple of lovely
interviews (the last one was a pair programming interview which I did at 9pm
from a family resort on not much sleep after a 4 hour drive) I got in!
<h2>What I'll be working on</h2>
<img src="/images/recurse-clip-cutter.png" alt="recurse-clip-cutter" />
I have a few projects in mind that I want to work on:
<ul>
<li>Finishing the <a href="https://github.com/jesusgollonet/opencv-clip-cutter">opencv clip cutter</a>, potentially looking into posture analysis for bmx and skate tricks</li>
<li>Looking into <a href="https://jesusgollonet.com/posts/watching-the-surf-in-vlc-with-streamlink/">webcam wave analysis</a></li>
<li>Finishing my <a href="https://github.com/jesusgollonet/website/tree/main/cli">blog cli tool</a></li>
</ul>
That said, just after getting access to the recurse platform, first thing I see
is an intro to <a href="https://supercollider.github.io/">supercollider</a> (which I was
really into in early 10s) and an algorave for tomorrow, so this is going to be
tough. I'm fully open to whatever comes my way.
<h2>Why I'm attending</h2>
There are a few reasons why I'm attending the recurse center:
<h3>More Programming!</h3>
In the last few years I've gotten a bit far from hands on programming. My work
has been more on the management and communication side, and when I was coding, I
was doing that more on the ops side, favoring serverless / managed solutions as
much as possible. There is nothing wrong with that approach and I will continue
to push for that, but that's closer to architecture, and it's making me lose a
building edge, which I miss.
<h3>More Community!</h3>
For the longest time, whenever I saw a blog post with an interesting deep dive
into some technical problem, I knew chances were high it came from a recurse
center alum. I want to hang out with those people!
<h3>Accountability</h3>
Related to community, having a group of people that can see what you're working
on is the best motivator to push yourself. I really want that in order to finish
some of the projects I have in mind.
<h3>Work prospects :)</h3>
I have been working in creative digital agencies for almost all my career. I
love that work, but I have always been curious about working in a more product
focused role. The recurse center offers a great opportunity to explore that
angle.
<h2>Logistics</h2>
I will be joining remotely from Spain. While I know attending in person in New
York would multiply the experience, I have been working remotely and across
timezones long enough to know how to make the most of it. Kids are on summer
holidays but my wife will be taking care of them in the afternoon /
evenings so I can focus on the program ❤️.
Can't wait to get started in a week!

I just got confirmed for the Summer 2 batch at the Recurse Center starting inJuly 1st and I'm through the moon.I have been wanting to attend for the longest time (I have tweets from as farback as 2...

I'm attending the recurse center!

In order to create an image with <a href="https://www.packer.io/">packer</a>, you need to define a template that
configures your image. Here' a simple example:
<pre><code class="language-hcl">source &quot;amazon-ebs&quot; &quot;example&quot; {
 access_key = var.aws_access_key
 secret_key = var.aws_secret_key
 region = var.region
 source_ami = &quot;ami-0c55b159cbfafe1f0&quot; # Amazon Linux 2 AMI
 instance_type = &quot;t2.micro&quot;
 ssh_username = &quot;ec2-user&quot;
 ami_name = &quot;packer-example {{timestamp}}&quot;
}

build {
 sources = [&quot;source.amazon-ebs.example&quot;]

 provisioner &quot;shell&quot; {
 script = &quot;provision.sh&quot;
 }
}

</code></pre>
This template references a file called <code>provision.sh</code>. This is useful to do
additional configuration on the image before it's created. For the sake of
example here's what a provision shell file could look like:
<pre><code class="language-bash">#!/bin/bash
echo &quot;Hello, World!&quot; &gt; /tmp/hello.txt
sudo yum update -y
sudo yum install -y httpd
sudo systemctl enable httpd
sudo systemctl start httpd
</code></pre>
Let's say we have both files under the same directory:
<pre><code class="language-bash">.
├── provision.sh
└── template.pkr.hcl

</code></pre>
In order to create an image you need to run the <code>build</code> command. When running
it, you need to inform packer how to find the template. Assuming you're running
<code>packer build</code> in the directory you have your template, you can do that in 2
ways:
<pre><code class="language-bash">packer build template.pkr.hcl
</code></pre>
or
<pre><code class="language-bash">packer build . 
</code></pre>
where <code>.</code> is the current directory.
This is all well and simple, but things start getting a bit confusing when you
start nesting directories.
Let's say you have the following directory structure:
<pre><code class="language-bash">.
├── image
│ ├── provision.sh
│ └── template.pkr.hcl

</code></pre>
if you now run the command from the parent directory:
<pre><code class="language-bash">% packer build image/template.pkr.hcl
Error: Failed preparing provisioner-block &quot;shell&quot; &quot;&quot;

 on image/template.pkr.hcl line 44:
 (source code not available)

 1 error(s) occurred:

 * Bad script 'provision.sh': stat provision.sh: no such file or directory
</code></pre>
Packer can't find the provision script. This is because packer is looking for
the <code>provision.sh</code> file in the directory where you're running the command, not
where the template is.
You could be tempted to fix this by modifying the path to the provision script
so that it's relative to wherever you're running your command from, but that
makes your script less portable by imbuing it with your current structure.
My usual hacky way to fix this is to simply <code>cd</code> into the directory where the
packer image is before running the command. I do that between parentheses so
that once the script is ran my shell goes back to the original directory:
<pre><code class="language-bash">(cd image &amp;&amp; packer build template.pkr.hcl)
</code></pre>
While writing this I learned that you can also solve it by using <a href="https://developer.hashicorp.com/packer/docs/templates/hcl_templates/path-variables">HCL path
variables</a>.
Specifically, you can use the <code>path.root</code> variable to reference the directory
where the template is.
<pre><code class="language-hcl">provisioner &quot;shell&quot; {
 scripts = [
 &quot;${path.root}/provision.sh&quot;
 ]
}
</code></pre>
Honestly, while this could seem more elegant, it seems less clear to me. Unless
you know what <code>path.root</code> is, it's not immediately clear what it's doing.
Even reading the documentation:
<blockquote>
path.root: the directory of the input HCL file or the input folder.
</blockquote>
it still takes me a few reads to understand what it is.
So yep, I'm happy I found a more official way to solve this issue, but I will humbly
stick to my hacky way, as I suspect it looks clearer to a bigger percentage of devs.

In order to create an image with packer, you need to define a template thatconfigures your image. Here' a simple example:source &quot;amazon-ebs&quot; &quot;example&quot; {  access_key    = var.aws...

Template paths in packer build

Yesterday, this <a href="https://it20.info/2024/5/a-framework-to-adopt-generative-ai-assistants-for-builders/">framework reflecting on where and when to adopt coding
assistants</a> caught my attention.
The full article is worth a read, here's the graph that summarizes it:
<img src="/images/coding-assistant-framework.png" alt="coding-assistant-framework" />
The quick summary is that the further a given task is from your skill level, the
more risk you're taking by delegating to a coding assistant. That doesn't mean
you should not do it, just that you need to take that risk into account.
Anyway, It caught my attention because I have scars.
At the beginning of the year I started developing an app using Rust. I had 0
previous experience with the language, but I have been interested for years and,
hey, we have ChatGPT and copilot! Learning curves are a relic of the past,
right?
During the initial weeks I was able to progress fairly quickly. Sure I sometimes
got in a loop of copy / paste from ChatGPT =&gt; editor and then compiler error message =&gt; ChatGPT,
but I always ended up making it work...
Until I didn't. Close to easter I was wrapping up an initial release of the project.
There was a feature that required a small refactor using threading and then I got into the classic loop but this time I kept spinning circles for hours, with ChatGPT suggesting progressively gnarlier looking type signatures until it gave up and got back to its initial suggestion.
The reality is that I had a barely functional understanding of concurrency and the Rust
borrow checker, so if ChatGPT couldn't help, I was out of luck.
I had to spend a significant portion of easter paying off the knowledge debt I had incurred, or
in framework terms, my risk line got really long, and I got burned.
<img src="/images/danger-zone.png" alt="danger-zone" />
LLMs can get you really far really fast, but you always need to keep your risk in check.
This is similar to the classic <a href="https://en.wikipedia.org/wiki/Circle_of_competence">circle of competence</a> advice for investing:
<blockquote>
Limit the investments you make in areas where you have little knowledge or experience.
</blockquote>
<img src="/images/circle-of-competence.png" alt="circle-of-competence" />
The other takeaway I found insightful is in the 'green area' of the graph.
<img src="/images/green-area.png" alt="green-area" />
You need to stay relatively close to your knowledge frontier in order to apply
assistants effectively. Tasks that are right below your competence level (the boost zone) are often great to delegate, as they might be within your means but can often be tedious to do.
But the most fun area is the learning zone, where something is right above your competence level and you can use the assistant to pull you up. A tricky thing here is that it requires being aware of what your actual competence is, which
might sometimes be a fun exercise for your ego.
Funnily enough this is exactly how I got out of my Rust rut. I kept going back
to basics until I found where I stood, then started building back up.
P.S.: <a href="https://x.com/jesusgollonet/status/1795473331908436133">I kept thinking about the
graph</a> and came to the conclusion that a single line would make it
clearer.
<img src="/images/single-line-twitter-me.png" alt="single-line-twitter-me" />
And <a href="https://x.com/mreferre/status/1795712647905820706">ben11kehoe had exactly the same idea.
</a>
<img src="/images/single-line-ben-kehoe.png" alt="single-line-ben-kehoe" />
So yep, be careful out there.

Yesterday, this framework reflecting on where and when to adopt codingassistants caught my attention.The full article is worth a read, here's the graph that summarizes it:The quick summary is that...

A Rust rut. Coding assistants and knowlege debt

Just wrapped up the first day of of <a href="https://jonthebeach.com"> J On The Beach </a>, a conference that takes place in my
hometown (8 years and counting) but I had never attended before.
The main question coming into day 1 was to test how bad an idea was to take 3
different workshops on 3 different topics which I have little experience with.
I did survive, and finished the day with more energy that I started, so I
guess <del>curb</del> boost your enthusiasm.
These are the 3 I attended:
<h2><a href="https://jonthebeach.com/workshops/containers/">Container Internals By Sean Scott</a></h2>
I don't do any fancy things with Docker, but I love low-level
unixy stuff, so I thought I would check this one out. I did learn quite a few
things. My favorite ones were:
<ul>
<li>a tangible understanding of a union / merge / overlay filesystem and how containers
leverage its layers to provide an isolated environment, showing how files can be
shadowed by files in the upper layers. Sean demonstrated that by mounting a
filesystem with a few different layers.</li>
<li><code>docker history</code> it will show you everything that has happened in the build
process (and why it's a good idea to include a step to build a new image out of
your finished image, so you minimize the risk of exposing sensitive information)</li>
</ul>
<h3>References</h3>
<ul>
<li><a href="https://github.com/oraclesean/jotb">Workshop Slides and Lab</a></li>
</ul>
Here are some materials I used to prep for the workshop. These formed the basis
of the material covered
<ul>
<li><a href="https://medium.oraclesean.com/visual-docker-introduction-f83baf4c0798">Visual Docker: Introduction</a></li>
<li><a href="https://medium.oraclesean.com/visual-docker-images-and-containers-1efb52d02d3c">Visual Docker: Images and Containers</a></li>
<li><a href="https://medium.oraclesean.com/visual-docker-images-are-immutable-1af06b242777">Visual Docker: Images are Immutable</a></li>
</ul>
<h2><a href="https://jonthebeach.com/workshops/p2p-chat/">P2P Chat By Rüdiger Klaehn</a></h2>
As much as decentralization and p2p has been co-opted by some sector of tech (i won't mention any
key term for fear of getting more spambots), it's still an interesting and
valuable technology. I loved seeing how Rüdiger claimed that type of
pragmatism explictly when motivating the workshop.
He spent a good portion of the workshop explaining their technical philosophy,
and the underlying networking concepts. I went home with a bunch of networking
items to explore further (QUIC, Nat hole punching, ALPN, ...).
Given that most of the participants (me included) were not very familiar with
rust (In my case I had also just broke my vim rust setup by installing <a href="https://github.com/fannheyward/coc-rust-analyzer"> Coc Rust Analyzer</a> 😢), the hands on part was more like a walkthrough of the code, but the
incremental demos were super helpful to understanding the concepts.
I've been using rust in the last few months and as hard as it sometimes is, this
seems like a great opportunity to continue with it.
<h3>References</h3>
<ul>
<li><a href="https://hackmd.io/@GMgEoX9mQWKh09CQ-JktgA/rkrfZgOzR">Workshop Slides</a></li>
<li><a href="https://github.com/n0-computer/iroh-workshop-jonthebeach">Iroh Worshop Source</a></li>
</ul>
<h2><a href="https://jonthebeach.com/workshops/datastax/">Build Your own RAG With Mustafa Esra</a></h2>
This was the more enterprisey / vendor-centric one, but I have been curious about AI Agents,
Langchain and Vector DBs for quite a while. Luckily it started with a good half
hour of theory. I still don't fully get how meaning is captured/derived from embeddings,
but I know a little bit more.
The workshop leveraged managed services for everything, which intiially bummed
me but it's amazing how much you can accomplish without touching a line of code.
I particularly loved <a href="https://streamlit.io/">Streamlit</a>, which I had never heard of before.
<h3>References</h3>
I checked this tutorial just before the workshop, and it helped with context.
<ul>
<li><a href="https://docs.datastax.com/en/ragstack/docs/quickstart.html">ragstack tutoral</a></li>
<li><a href="https://github.com/michelderu/ragstack-astradb">source for the workshop</a></li>
</ul>
<h2>That's it!</h2>
It's been quite interesting to experience the different types of workshops, but
this is going to be it for now. Let's see what the full conference has to offer.

Just wrapped up the first day of of  J On The Beach , a conference that takes place in myhometown (8 years and counting) but I had never attended before.The main question coming into day 1 was to te...

J On The Beach: Workshops

A few days ago I saw this video about <a href="https://www.surfline.com/">Surfline</a>'s <a href="https://www.youtube.com/watch?v=xecRZIM6Vlg">LOTUS Surf Forecasting System</a>.
They have hundreds of webcams around the world, and they analyze the streams to understand and predict the quality of the waves.
<a href="https://www.youtube.com/watch?v=xecRZIM6Vlg"><img src="/images/surfline-lotus.png" alt="surfline-lotus" /></a>
Surfline does not seem to have any webcams where I live (Southern
Mediterranean), which is understandable, given the questionable quality of the surf.
But the beach is still a big attraction, and there are plenty of webcams pointed at it.
I've been <a href="https://github.com/jesusgollonet/opencv-clip-cutter"> playing with computer vision
</a> lately and want to give some basic wave detection a try. The first step is capturing streams from beach webcams, but they usually appear in sites which are slow and full of ads.
<a href="https://streamlink.github.io/">Streamlink </a>
is a great tool for that. It allows you to watch streams in VLC and provides a Python API that gives access to the stream data.
If the stream you care about is in a popular platform, like twitch or youtube
(or many others!) chances are there is <a href="https://streamlink.github.io/plugins.html">streamlink plugin</a> for it. Usage in this case is as simple as
<pre><code class="language-bash">streamlink https://www.twitch.tv/merewether_slsc best 
</code></pre>
If there is no plugin, you can look for the <code>m3u8</code> file in the page source and pass it to streamlink.
<pre><code class="language-bash">streamlink 'https://hd-auth.skylinewebcams.com/live.m3u8?a=rniaof33h8kqu8muddr7t086v3' best 
</code></pre>
Some sites do have some protection against this, but it hasn't been too hard to get around it.
Streamlink allows you to pass cookies and headers to the request, so what I've done to figure out what headers were needed was to copy the curl
request from the network panel and tweak them so they have the correct streamlink format for the <code>--http-header</code> option.
<img src="/images/image_2024-04-30-11-22-11.png" alt="image_2024-04-30-11-22-11" />
In the couple of sites I tried, it has been enough to simply pass the origin
header to streamlink, so from
<pre><code class="language-bash">curl 'https://hd-auth.skylinewebcams.com/live.m3u8?a=rniaof33h8kqu8muddr7t086v3'
...
 -H 'origin: https://www.skylinewebcams.com' \
...
</code></pre>
to
<pre><code class="language-bash">streamlink --http-header 'origin=https://www.skylinewebcams.com' 'https://hd-auth.skylinewebcams.com/live.m3u8?a=rniaof33h8kqu8muddr7t086v3' best 
</code></pre>
<img src="/images/image_2024-04-30-11-25-04.png" alt="image_2024-04-30-11-25-04" />
And I will leave it at that, as I just saw there might be some waves to catch
today! 🏄

A few days ago I saw this video about Surfline's LOTUS Surf Forecasting System.They have hundreds of webcams around the world, and they analyze the streams to understand and predict the quality of th...

Watching the Surf in VLC with Streamlink

This blog is handcrafted with markdown files in vim. Up until now, posting
images was a prohibitive dance of putting the image on a certain folder and
figuring out the relative path, posting it into markdown and then adjusting a
few times until I got it right.
Which is the reason why I never posted any.
Yesterday I implemented <a href="img-paste.vim">https://github.com/img-paste-devs/img-paste.vim</a> <a href="#user-content-fn-1" id="user-content-fnref-1" data-footnote-ref="" aria-describedby="footnote-label">1</a>
and now I can paste images directly.
If you see the below image, it works!
<img src="/images/image_2024-04-25-13-52-54.png" alt="image_2024-04-25-13-52-54" />
<h2>Demo</h2>
Now the steps to add an image to a post are:
<ul>
<li>Copy the image to the clipboard: I usually use <code>Cmd+Ctrl+Shift+4</code> on a mac</li>
<li>Go to the markdown file in vim</li>
<li>Press <code>&lt;Leader&gt;p</code> to paste the image (my leader key is comma, so in my case, <code>,p</code>)</li>
</ul>
This will create a png file in the directory you specify, and add the right
markdown syntax to the file.
Here is a quick video explaining how it works:
<a href="http://www.youtube.com/watch?v=QAdq7jXObEo"><img src="http://img.youtube.com/vi/QAdq7jXObEo/0.jpg" alt="A video of how i paste and preview images in markdown using vim" /></a>
<h2>Some Notes</h2>
Instructions on the plugin page are clear. Here are a couple of adjustments I
had to make to match to my workflow
<h3>Correct the image path</h3>
By default, the plugin stores the image in an <code>img</code> subdirectory inside the same folder
as the markdown file. In my case I wanted to move it to my /public/images folder at the root of my
site.
<pre><code class="language-vim">let g:mdip_imgdir = '../public/images'
</code></pre>
I have added that to <a href="https://github.com/jesusgollonet/website/blob/develop/.vimrc.local">a <code>.vimrc.local</code> file In the root of the project</a>, since this
configuration is specific to this project.
<h3>Rewrite URLs</h3>
The above variable creates a url that is relative to the markdown file, which is
perfect for when you're previewing the markdown file with an external tool or
withing github.
But in Next.js (which is what I use for this website), the /public folder is served as
the root of the website (so /public/images is just /images).
To solve this, I added <a href="https://github.com/jesusgollonet/website/blob/7f6047d94e824fc9c3087b986c830a11de339c01/build/lib/posts.ts#L25C1-L31C5">a simple regex that rewrites the links when rendering</a>
<h3>Further</h3>
Here's a few improvements I'm thinking of:
<ul>
<li>The obvious next step is to add the ability to make it work with image paths to
avoid having to only do it through the clipboard.</li>
<li>More in general, I would love to be able to paste a url and get the markdown
tags automatically, ideally being smart enough to know if it's a url, a local
file or even a youtube embed (TIL <a href="https://github.com/adam-p/markdown-here/wiki/Markdown-Cheatsheet#videos"> there's no good way to embed youtube videos </a>)</li>
</ul>
And that's it. If you have comments questions or suggestions please let me know!
<a href="https://twitter.com/jesusgollonet">@jesusollonet</a> on twitter (or pretty much
anywhere) or email me at here.
<section data-footnotes="" class="footnotes"><h2 id="footnote-label" class="sr-only">Footnotes</h2>
<ol>
<li id="user-content-fn-1">
Oh god I had to add the markup to this link manually!! Next one. <a href="#user-content-fnref-1" data-footnote-backref="" aria-label="Back to reference 1" class="data-footnote-backref">↩</a>
</li>
</ol>
</section>

This blog is handcrafted with markdown files in vim. Up until now, postingimages was a prohibitive dance of putting the image on a certain folder andfiguring out the relative path, posting it into m...

Pasting images in Markdown using Vim

As I have started getting back into the rhythm of coding more or less daily, I have been paying
attention to my workflow. My favorite piece on the topic is Simon Willison's masterful
<a href="https://simonwillison.net/2022/Jan/12/how-i-build-a-feature/"> How I build a feature </a>, which provides a detailed guide on building software including everything from tests to documentation and publishing.
While I'm very far from such heights, I have been almost invariably applying his first point:
<blockquote>
Everything starts with an issue
</blockquote>
I'm mostly applying this to personal projects, but even then using github issues
has proven extremely helpful:
<ul>
<li>They serve as TODO list for me. Especially handy when applied on projects that
I pick up every few weeks/months.
e.g:</li>
</ul>
<pre><code class="language-bash">gh issue list

ID TITLE LABELS UPDATED
#50 content: add /now page about 26 days ago
#49 cli: prompt save or discard post after editing about 2 months ago
#48 cli: check if server is running and optionally start about 3 months ago
#47 investigate how to share lib/posts between cli and website about 3 months ago
#46 fe: add syntax highlighting to inline code blocks about 3 months ago
#45 fe: break down diary page into components about 3 months ago
#39 cli: load configuration from entrypoint instead of inside new-post about 4 months ago
#33 cli: list out features + mvp about 4 months ago
#21 Simple web editor about 10 months ago
</code></pre>
<ul>
<li>I can add extra context in the form of comments (e.g: when finding relevant
links for a given topic).</li>
<li>They add a fun touch of gamification to my work.</li>
</ul>
In parallel, I have been using the <a href="https://cli.github.com"> gh cli </a> more and
more, and have stumbled upon a workflow that I find almost soothing in its
assemply line quality. The steps are pretty obvious. For every feature (or problem, bug, investigation...), I:
<ul>
<li>create a new issue</li>
<li>work on issue, and eventually fix it.</li>
<li>create pr</li>
<li>merge pr</li>
</ul>
Each of the steps maps perfectly to a gh command:
<ol>
<li>
Create an issue in github:
<pre><code class="language-bash">gh issue create -t 'Fix scroll in the homepage'
</code></pre>
</li>
<li>
Start working on it
<pre><code class="language-bash">gh issue develop &lt;issue number&gt; -c
</code></pre>
This is my favorite one. The <code>develop</code> command creates a branch named like the issue id and title. E.g: (e.g: <code>12-fix-scroll-in-the-homepage</code>). You can get the issue number using <code>gh issue list</code>. The <code>-c</code> flag means checkout the branch.
Commit, push, commit, push... done?
</li>
<li>
When the issue is ready, I'll create a PR. I will confess this part is superfluous when working on my own. I don't add any extra details to the PR, and seldomly review the changes, but I still do it in order to instill good practices for team settings.
<pre><code class="language-bash">gh pr create -t &quot;my pr title&quot;
</code></pre>
</li>
<li>
Then merge I the branch and delete de PR and branch.
<pre><code class="language-bash">gh pr merge &lt;pr number&gt; -r -d
</code></pre>
<code>-r</code> is for rebase and <code>-d</code> deletes the branch (local and remote) after merging.
</li>
</ol>

As I have started getting back into the rhythm of coding more or less daily, I have been payingattention to my workflow. My favorite piece on the topic is Simon Willison's masterful How I build a fe...

gh cli zen workflow

I have started a project in Python, which I'm not particularly familiar with. I
thought it would be useful to include in the git repo relevant shell commands to
remember what I've done, and then wondered if it was possible to have shell history
but limited to a specific directory.
Turns out this is trivial with <a href="https://github.com/ellie/atuin">atuin</a>, which
records your shell history to a sqlite database and provides additional context.
Assuming you're in the directory you want to check your history for:
<pre><code class="language-bash">$ atuin search --cwd $(pwd) term
</code></pre>

I have started a project in Python, which I'm not particularly familiar with. Ithought it would be useful to include in the git repo relevant shell commands toremember what I've done, and then wonde...

Directory specific shell history

For the next 8 weeks, I'm going to follow an extremely minimal strength training
plan of 4 exercises:
<ol>
<li>Pull ups</li>
<li>Split squats</li>
<li>Push ups</li>
<li>Step ups</li>
</ol>
I'm going to do 3 sets for each, pairing 1-2 and 3-4. The number of reps is
still being figured out (I'm 2 weeks in), but anywhere from 5 to 15, depending
on the exercise. Since I don't have weights at home, I'll increase the rep range
by 5-10% per week.
In total I'm aiming to being done in under 25 minutes, including 5 min warmup,
and if there's any time left I'll do a jump rope finisher.
This is extremely small and and simple. And even though the exercises are
reasonably full body, it no doubt leaves stuff on the table. The
reasons why I still think it's worth comitting to are:
<ul>
<li>I still need to make training a habit. I've started many times but there's
always something that gets me back to the starting line. This is small enough
that I can't say no to under almost any circumstance.</li>
<li>I need to build basic strength before getting into fancier exercises. There's
no shortage of Skate/Surf specific content on instagram and youtube, but after
trying a few things, it's clear to me that I need a solid foundation.</li>
<li>As a fulltime working dad, that also wants to skate and BMX and surf (and many
other things!), I need to be extremely smart with time management.</li>
<li>It can't interfere with random bmx/skate sessions throughout the week. There's
nothing more frustrating than not being able to do your sport because you're too
sore from training.</li>
<li>I don't want to rely on a gym. Apart from pull ups. The rest of
exercises can literally be done anywhere.</li>
<li>Having fewer exercises and low volume should allow me to focus on movement
quality.</li>
<li>Any plan is better than no plan, and this is mine.</li>
</ul>

For the next 8 weeks, I'm going to follow an extremely minimal strength trainingplan of 4 exercises:Pull upsSplit squatsPush upsStep upsI'm going to do 3 sets for each, pairing 1-2 and 3-4. Th...

Minimally Viable Plan (Strength Training)

I have missed a couple of yoga sessions in the last 4 days. I missed last week's
weekly post and am late for this week. I was starting to question why even
write? I don't have much to share...
Which is what's on my mind, and isn't this about sharing exactly that?
So that's a post.

I have missed a couple of yoga sessions in the last 4 days. I missed last week'sweekly post and am late for this week. I was starting to question why evenwrite? I don't have much to share...Which i...

Back to blocked?

I have a bunch of <a href="https://github.com/jesusgollonet/dotfiles/blob/master/zsh/aliases"> alliases in my shell
</a>. When
coding in vim, I frequently run whatever file I'm working on from vim's command
mode,
<pre><code class="language-vim">:!node %
</code></pre>
often by just mapping a key combo to run it.
<pre><code class="language-vim">:map ,r :w\|!node %&lt;CR&gt;
</code></pre>
A problem I often run into is that vim's command mode
doesn't know about my shell aliases. So if I have an alias like this in .zshrc
<pre><code class="language-bash">alias python=python3
</code></pre>
When running <code>:!python %</code> from vim, I get something like
<pre><code class="language-vim">zsh:1: command not found: python

shell returned 127
</code></pre>
My aliases used to be defined in .zshrc, but this file is not sourced when
running a command from vim's command mode. The solution is to <a href="https://github.com/jesusgollonet/dotfiles/blob/master/zsh/zshenv">move your
aliases to <code>~/.zshenv</code>
</a>, which is
sourced.
Note there's no need to load ~/.zshenv from ~/.zshrc, as it's part of the <a href="https://zsh.sourceforge.io/Intro/intro_3.html">
default zsh startup sequence </a>
and therefore loaded by default.

I have a bunch of  alliases in my shell. Whencoding in vim, I frequently run whatever file I'm working on from vim's commandmode,:!node %often by just mapping a key combo to run it.:map ,r :w\|...

Shell aliases in vim command mode

Even later than usual. Was thinking of skipping this week, but my Gestalt sense
didn't let me sleep (good job on labelling the weeklies with successive numbers).
I have continued to record sessions and have started uploading my quick edits to <a href="https://www.youtube.com/@jesus_gollonet"> youtube </a>.
So far I've uploaded a couple of raw sessions with empty / non interesting parts
remmoved, and I love having the ability of checking them whenever / wherever.
The main goal is to check and help with progression. But I also think video
editing is a very useful skill to have.
Regarding that, iMovie is both great and terrible. It's at hand, and you can
slice clips in no time, but then you want to do simple things like a slight
rotation to correct horizon and you need to open Keynote (!!!). Have been
checking out Filmora and it seems very promising.
I've been thinking about video analysis, I have the feeling it can be
tremendoussly useful, even without specialized biomechanics knowledge. I haven't
found a free app to test it out, but thinking about the features (load and
display a couple of videos side by side, play/pause/skip forward, slow motion,
draw simple lines on top), it sounds like something so apt for openframeworks it
could even be a beginner tutorial. I have downloaded xcode and oF for the first
time in many years. I might give it a try if I can find time.
20-30 min of Yoga continues to be my only workout other than skate/bike. Last
Friday was the first time I've skipped since I started, but only because I
fell asleep on a Yin Yoga session after dinner and drinks at 3am.
Save, publish, and Yoga. Until next week.

Even later than usual. Was thinking of skipping this week, but my Gestalt sensedidn't let me sleep (good job on labelling the weeklies with successive numbers).I have continued to record sessions an...

Week 4

I had to chuckle at this:
<blockquote>
...you really can drive to the skatepark and it’s like a gym
</blockquote>
<blockquote>
...you primarily have skateboarders who come from homes where their parents would drive them to a sporting environment.
</blockquote>
From this <a href="https://www.skateism.com/ben-koppl-hates-being-famous-online/">interview with Ben
Koppl</a>, my
favorite skater which unsurprisingly seems to be an interesting person. He's
talking about the changing demographics of skateboarding.
I feel strongly that treating skateboarding more like a regular sport is a
great thing (it's the reason why I'm even considering it at 46!), and I'm 100% a
parent that drives his kids to the skatepark, but I love his point about
skateboarding starting as a mode of transportation, and a way for kids that have
nothing else to do or want to avoid their neighbourhood to have a tool to do
that while pushing their creativity...
The full interview is interesting, I enjoyed his non-negative take on kids
perferring short instagram clips vs video parts (which I've often seen mourned
about in BMX circles), and the early youtube days.

I had to chuckle at this:...you really can drive to the skatepark and it’s like a gym...you primarily have skateboarders who come from homes where their parents would drive them to a sporting env...

Parents that drive their kids to a skatepark

Ok let's get this published on <del>the Monday</del> Tuesday.
So now I have a problem, whenever I go to a skatepark, I have to choose between
the skateboard and bike. So far I've been basically splitting the time based on
what I feel like doing, often switching every 20 minutes. Trying to progress on
both is going to make it slower but oh well.
Been early to a public skatepark a couple of mornings, setting up the tripod and
sessioning some things. I've already spotted a couple of bad habits from that
(rotated shoulders when ollieing, butt not back enough on manuals on the
bike...), so plan to do it more.
I'm also looking for fun ways to put those into mini edits. The idea is similar
to what I'm doing here. Capture quickly, edit so that it's not super boring and
always publish. <a href="https://www.instagram.com/p/CsjGAF3Is2L/">Here's one</a>
Inspired by <a href="https://www.instagram.com/p/CsEcOPjNfix/">the daily push flick analysis
videos</a> I started researching better
ways to do video analysis, and daydreaming about automating / AI-ing some parts
of the process. Maybe
On the workout front, Yoga continues, and I started doing 5 minutes of jump rope
(which I've stopped for a few days as I hurt my knee on Sunday). I'm listening
again to the <a href="https://hubermanlab.com/dr-andy-galpin-how-to-assess-improve-all-aspects-of-your-fitness/">Huberman and Andy Galpin Fitness assessment
episode</a>
and plan on doing some assessment this week.
That's mostly it, skating, recording and doing informal analysis, and
obsessing over fitness.
Talk soon.

Ok let's get this published on the Monday Tuesday.So now I have a problem, whenever I go to a skatepark, I have to choose betweenthe skateboard and bike. So far I've been basically splitting the tim...

Week 3

Oh my god. Wednesday!
Last week was slower in all fronts.
Finished my 3 weeks of 20 minutes of Yoga a day. Last couple of days it
got really late and I did Yin Yoga instead. I'm resisting calling it 'cheating',
as I've invariably found it helpful when I've done it in the past, and I like to
think of myself as a strong proponent of less Yang more Yin, but yep, that is a
thought that has been had..
I have been going back and forth with what to do next, and I still don't have a
decision. I'm low-key terrified of losing momentum, as it is so hard to return.
I also want to start doing more, for some definition of more, but always in a
way that doesn't risk breaking the habit. Continuing Yoga for the moment.
Had the first couple of sessions on a skateboard. Taking it easy as it's
significantly tougher on joints and tendons, but I'm glad with where I'm at.
Managed to drop in on a couple of ramps without giving it 2 seconds, which was
really encouraging. It's been at least 10, if not 12 years since I last skated,
Couple of shy starts at Python. Getting past the environment setup/conventions,
and looking forward to starting the fun part.
Anyway, Save, publish... here's to a better one.

Oh my god. Wednesday!Last week was slower in all fronts.Finished my 3 weeks of 20 minutes of Yoga a day. Last couple of days itgot really late and I did Yin Yoga instead. I'm resisting calling it '...

Week 2

I'm going to start a weekly recap to keep writing/publishing momentum. This
is written in haste as a quick capture of things I'm up to so don't expect too
much.
I decided to extend the 20 min of daily yoga to 3 weeks, and haven't missed a
day so far. Duration is low enough that I don't need any rest days, and even if
I'm tired I can get myself to do it. I'm also using instagram stories as
outsourced accountability and that's been working wonders.
The main impulse I've struggled with controlling is adding more time, more exercises, starting
cardio... Doing too much too soon is the killer of any fitness initiative so
I'll keep the commitment to 20 min.
Interestingly, even with such small time I can already feel some benefits. I was
at the skatepark early in the week and then over the weekend trying similar
tricks and the second time I could definitely feel better strength and
stability. It could well be something else but I wouldn't be surprised if the
yoga already has an effect.
I got a tripod and have started recording sessions at the skatepark. Curious to see if it can
help me progress better, as the few times I've recorded myself in the past have
demonstrated just how different things feel in your head vs from the outside
(and you can guess which one is nicer).
I also got a skateboard! That used to be my main thing growing up, and I re-picked it
up ocassionally in BCN (late twenties) and Amsterdam (mid thirties). I hadn't
considered it in the last few years due to some knee and ankle issues, but a friend lent me one for 10
minutes a few days ago and I could feel way more confident than last time I
tried 2 years ago, so got a super cheap second-hand complete and plan to
take it slowly.
Damn it's Tuesday already, so here's a list of a few other things I'm up to:
<ul>
<li>Using <a href="https://www.instagram.com/jesusgollonet/"> instagram as a raw visual diary </a>, and deliberately capturing stuff I used to be shy/embarrassed about.</li>
<li>Starging to feel a bit more comfortable with fakie on the BMX (and recording!)</li>
<li>Cooking quite a bit (and recording!)</li>
<li>After getting burned by <a href="https://twitter.com/jesusgollonet/status/1652386580387045376">a stupid typescript rabbithole</a> I <a href="https://github.com/jesusgollonet/website/blob/main/build/posts/2023-04-30_a-typescript-rabbit-hole.md">shelved a blog entry</a>
and haven't writen anything here, which is one reason why I'm starting this
weekly series.</li>
<li>After a few mindblowing experiences with ChatGPT I've finally tarted to dive
properly into AI. The <a href="https://www.deeplearning.ai/courses/#short-courses">prompt engineering
course</a> is proving super
valuable.</li>
<li>Relatedly, it seems like the <a href="https://twitter.com/jesusgollonet/status/1654536311661797384"> time finally came
</a> for me to get
into Python. A lot of the material on openAI seems to have decent Typescript or
js support but the native tongue is definitely Python, so there we go.</li>
</ul>
And there's more. Seems like there's a lot of momentum in many areas, so let's
see how long we can keep it.
Talk soon!

I'm going to start a weekly recap to keep writing/publishing momentum. Thisis written in haste as a quick capture of things I'm up to so don't expect toomuch.I decided to extend the 20 min of daily...

Week 1

The hardest part of exercising is getting back to it once you inevitably fall off
the wagon.
I started the year on a good streak of almost 2 months doing
reasonably regular exercise, including 2 to 3 days a week of strength and
at least one of low HR cardio. But then I got a new job, a stressful first
project and some other life events. So it's been a couple of months off again.
I have not completely stopped moving. I've been hitting the skatepark <a href="https://www.instagram.com/p/CrftUXQIhgZ/">on
a BMX with my kids</a> almost every week, but as I'm learning, the consistency
and focus of an actual workout is the key to improving, especially well into your 40s <a href="#user-content-fn-1" id="user-content-fnref-1" data-footnote-ref="" aria-describedby="footnote-label">1</a>.
There's the part where you need to get motivated enough to start again, but even
harder is the part where you need to re-adjust your expectations. You
were able to do X pull-ups, run Y kilometers, and now it's less.
So I've started with low expectations. 20 minutes of yoga every day for a week.
I need to work on a bit of low back pain and just get mobility in order, then
I'll get back to strength and cardio.
This is day 3. 🤞
<section data-footnotes="" class="footnotes"><h2 id="footnote-label" class="sr-only">Footnotes</h2>
<ol>
<li id="user-content-fn-1">
Beth Lewis said something in <a href="https://www.youtube.com/watch?v=t48iqNwS6PY">The Art of Stability</a> that stuck with me:
<blockquote>
Don't let your sport get in the way of your training.
</blockquote>
<a href="#user-content-fnref-1" data-footnote-backref="" aria-label="Back to reference 1" class="data-footnote-backref">↩</a>
</li>
</ol>
</section>

The hardest part of exercising is getting back to it once you inevitably fall offthe wagon.I started the year on a good streak of almost 2 months doingreasonably regular exercise, including 2 to 3 ...

Resuming exercise

I have been meaning to get back to writing in public for many years, but I always get blocked at different stages:
What engine do I use? Do I host myself or do I go for a managed option? Should I keep it just professional or is it ok to get personal? Single blog for all my interests or a few thematic ones? And &quot;who would even care&quot;? etc.
When the urge comes back, I start making decisions but then I run out of time
and let it sleep for another few months.
And it's been checks notes 8 years since <a href="https://github.com/jesusgollonet/software-over-the-rainbow/blob/master/build/source/_posts/2015-04-21-this-undefined-es6-arrow-functions.markdown">I last published an
entry</a>
but more like 12 years since the last time I blogged
with any regularity (here's <a href="https://github.com/jesusgollonet/software-over-the-rainbow/tree/master/build/source/_posts">the archives of my old blog</a> ).
I've had some periods of activity on <a href="https://twitter.com/jesusgollonet">Twitter</a> and <a href="https://www.instagram.com/jesusgollonet/">Instagram</a> (I'm currently loving it!) and even tried <a href="https://dev.to/jesusgollonet">a different platform</a>, but I'd like to have a place where I can explore freely.
I'm not getting blocked this time. I'm not even going to edit this one before
publishing. Wait what was the way to deploy this again?
If you're reading this, I managed to figure it out :)

I have been meaning to get back to writing in public for many years, but I always get blocked at different stages:What engine do I use? Do I host myself or do I go for a managed option? Should I keep...

Diary