México 🇲🇽
Joined September 2011
- Tweets14.8K
- Following434
- Followers904
- Likes19.9K
Now that makes me want to see @Delta_Emulator take advantage of the Duo to really recreate the Nintendo 3DS experience.
Mathieu Céraline retweeted
13 years ago, Apple asked a question that hits me harder today than ever:
If everyone is busy making everything, how can anyone perfect anything?
Every day, I see a world flooded with effortless AI slops. I’m realizing that to build something truly great today, you need a level of focus the world has never demanded before.
It feels like trying to write a quiet, beautiful song in a room full of loud noise.
At least, that will be my personal bet over the next few months.
I want to minimize + become extremely knowledgeable and aware of every part of my stack.
Being dependency free is unlikely, but in a world where “code is cheap” and knowledge is plenty, I want to “own” the things I depend on the most, at least when it makes sense.
With the rise of cyber and supply chain attacks, I feel like the future is in “highly personal software” (at least when it makes sense).
Software, highly customized for the user, either with pinned dependencies or without dependencies altogether, decentralized, and away from the prying eyes of the giants (and ideally the internet).
My first step toward achieving that… building my own obsidian plugins!
I use Obisidian more than a dozen times everyday and I rely on a handful (20+) plugins that have become pretty essential to my workflows, yet:
- They often are either a bit short (or large) of what I need.
- I never met the makers, yet clicking the “update” button requires a lot of trust.
I am currently working on two plugins that I hope to make public soon:
1) Depot: a S3 compatible file uploader that supports private buckets
2) Shard: a command palette that supports smart search and actions (imagine writing “games create The Witcher 3” in the search bar instead of clicking a dozen buttons)
For other plugins that I consider “complete” (meaning their feature set match my needs and I don’t anticipate requiring updates for), I will either pin their version or clone + make a private plugin with them, only for myself, depending on usecases.
With the rise of cyber and supply chain attacks, I feel like the future is in “highly personal software” (at least when it makes sense).
Software, highly customized for the user, either with pinned dependencies or without dependencies altogether, decentralized, and away from the prying eyes of the giants (and ideally the internet).
My recipe to be the most efficient with AI is… opinions.
Over the last few months, on internal or client projects, I have been refining clearly how I like to work, in an attempt to be able to discriminate very fast what is good or bad code at a glance.
Since I became more opinionated and knowledgeable about “how we do things around here”, it is a lot easier:
- for me to direct the agents in the right direction from the start
- for agents to be autonomous, do and review their work
- for me to manage multiple projects, context switching or going back to older projects
- for me to read the code, review it and ship it with confidence
This is my bet for the future with @madincohq. We live in a tech bubble so it’s easy to think that everyone already figured it all out with AI, that everyone is already building their own software factory etc.
But outside the X bubble, there are so many small businesses owners, non profits, mom and pop’s stores that are yet to do their digital transformation and that need people to help them navigate and take advantage of this new world.
With Madinco, overtime, I wish to empower all of these people. Help them do more with technology.
To change pace and stop talking about tech for a bit…
1) What’s your favorite video game?
2) What’s your favorite TV show?
3) What’s your favorite book?
Even with AI, it still so hard to make "good" software.
I already have a mostly complete version of Depot, but I am still sweating the details. Even if few will ever care to read the code, I still want it to feel intuitive, maintenable and well optimized.
Software has become easy, but well crafted software is as hard as ever.
So far, here is how it has shaped up:
- 🔒 Keep images and files in your own private S3 or R2 bucket
- 🔗 Notes store a clean mybucket://photo.jpg link (mybucket can be the name you want): no public URLs, no expiring tokens
- 📋 Paste, drag and drop, or pick a file
- 🖼️ Images show up in Reading view, Live Preview, and Bases card covers
- 📄 PDF, audio and video embeds
- ⚡ Cached on your device, so images load instantly and work offline
- 🗂️ You can have multiple storage profiles (personal, work, public…)
- 🛤️ You define the file path: by date, by type, by note name
- ♻️ Duplicate upload detection (optional)
- ✏️ If you rename a path, it detects and optionally updates every link in your vault in one go
- 🖼️ A library view: browse everything in your bucket, search it, and see what's used, unused or broken
- 🔑 Credentials stay in Obsidian's secret storage, never in your notes or in the plugin
With only one dependency to sign the requests.
Definitely not yet ready to be published, I will continue crafting for a bit, but quite exciting so far!
Funnily (or sadly) enough, the biggest drawback of Obsidian is his greatest advantage: it's just text files on the filesystem.
There is no easy way to protect our notes from prying eyes, whether humans, or rogue agents/background processes, etc.
With everything going on these days, the so-called “cyberpocalypse”, society might end up putting less and less trust in centralized systems.
At the same time, with AI, we are very likely to see the emergence of ultra personalized software or at the very least smaller centralized silos.
Not just by developers and engineers but by everyone, which is a double edge sword.
On the one hand, less technical people building apps, even with hyper intelligent LLMs… is bound to cause issues. At least for the time being, creating great software is still very hard, especially when it comes to security, performances, etc.
On the other hand though, less centralized systems means more variety in the attack targets, from a cybersecurity standpoint. So even less safe platforms might have a better survival chance even is just for the sake of diversity.
Desperately want to try out @Cloudflare’s durable object as an attempt to better understanding their whole offering and what I can leverage in my projects… but I haven’t quite found a usecase for DO yet. 😅
In the spirit of this little experiment, after it has shipped, and when I will have some time, I would love to try making my own theme for the first time.
I love minimal, border, and many themes. I love flexoki too. But I feel like there many small and subtile changes I would like to make for them to better fit what I want.
So far, here is how it has shaped up:
- 🔒 Keep images and files in your own private S3 or R2 bucket
- 🔗 Notes store a clean mybucket://photo.jpg link (mybucket can be the name you want): no public URLs, no expiring tokens
- 📋 Paste, drag and drop, or pick a file
- 🖼️ Images show up in Reading view, Live Preview, and Bases card covers
- 📄 PDF, audio and video embeds
- ⚡ Cached on your device, so images load instantly and work offline
- 🗂️ You can have multiple storage profiles (personal, work, public…)
- 🛤️ You define the file path: by date, by type, by note name
- ♻️ Duplicate upload detection (optional)
- ✏️ If you rename a path, it detects and optionally updates every link in your vault in one go
- 🖼️ A library view: browse everything in your bucket, search it, and see what's used, unused or broken
- 🔑 Credentials stay in Obsidian's secret storage, never in your notes or in the plugin
With only one dependency to sign the requests.
Definitely not yet ready to be published, I will continue crafting for a bit, but quite exciting so far!
Agreed, less is more. Though achieving “less” is the hardest part, especially in an interconnected world.
I would also add:
- defining strategies and routines/SOPs for things like updating dependencies, OS, runtimes, etc.
- generally being more knowledgeable about the dependencies we do choose to have, even in our personal lives
- using ssh agents and managing secrets well (not just leaving them in an .env file)
- disabling automatic updates when relevant
- enabling very frequent backups when relevant
- generally identifying risks, choosing what we want to manage or delegate, and having contingencies in place for when things *do* go wrong.
Instead of cookie banners no one cares about, I think the @EUCouncil should really focus on defining and enforcing standards for “offboarding” from internet platforms.
Many still don’t offer the ability to delete your account, let alone downloading all of your data before doing so. And even for those that do, we need a standard:
- To actually end up with usable data after you offboarded
- Delay in treatment for deletion request, and some way to verify it has been done
The internet of the last 30 years was based on trust and that’s how it grew.
No one owns the full supply chain, we all just “take a little bit of this here and there”, with OSS, with PaaS and SaaS, with npm/composer, SSL certificates, analytics providers, etc.
But this is actively changing. Now no one can be trusted. It’s not just about sanitizing user inputs anymore, we need to be actively aware of EVERY part of our supply chain and, to some extent, have contingencies in place.
On a personal level it is also hard to clean up, I have been spending the last year occasionally requesting deletion of my data on various services I used, where I have accounts, etc. It’s crazy how much our data is sprawling a bit everywhere and how many “dependencies” we end up having.
Consolidating/minimizing dependencies is possible but takes real real effort and time. But I think it’s necessary.