Full Width [alt+shift+f] Shortcuts [alt+shift+k]
Sign Up [alt+shift+s] Log In [alt+shift+l]
102
It’s been around 2 years that I’ve had to stop with my long-term addiction to stable jobs. Quite a few people who read this blog are wondering what the hell exactly I’ve been doing since then so I’m going to update all of you on the various projects I’ve been working on. Meme credit: Fabian Stadler Mikochi Last year, I created Mikochi, a minimalist remote file browser written in Go and Preact. It has slowly been getting more and more users, and it’s now sitting at more than 200 GitHub stars and more than 6000 Docker pulls. I personally use it almost every day and it fits my use case perfectly. It is basically feature-complete so I don’t do too much development on it. I’ve actually been hoping users help me solve the few remaining GitHub issues. So far it happened twice, a good start I guess. Itako You may have seen a couple of posts on this blog regarding finance. It’s a subject I’ve been trying to learn more about for a while now. This led me to read some excellent books including...
9 months ago

Improve your reading experience

Logged in users get linked directly to articles resulting in a better reading experience. Please login for free, it takes less than 1 minute.

More from Alice GG

Playing with open source LLMs

Every 6 months or so, I decide to leave my cave and check out what the cool kids are doing with AI. Apparently the latest trend is to use fancy command line tools to write code using LLMs. This is a very nice change, since it suddenly makes AI compatible with my allergy to getting out of the terminal. The most popular of these tools seems to be Claude Code. It promises to be able to build in total autonomy, being able to use search code, write code, run tests, lint, and commit the changes. While this sounds great on paper, I’m not keen on getting locked into vendor tools from an unprofitable company. At some point, they will either need to raise their prices, enshittify their product, or most likely do both. So I went looking for what the free and open source alternatives are. Picking a model There’s a large amount of open source large language models on the market, with new ones getting released all the time. However, they are not all ready to be used locally in coding tasks, so I had to try a bunch of them before settling on one. deepseek-r1:8b Deepseek is the most popular open source model right now. It was created by the eponymous Chinese company. It made the news by beating numerous benchmarks while being trained on a budget that is probably lower than the compensation of some OpenAI workers. The 8b variant only weights 5.2 GB and runs decently on limited hardware, like my three years old Mac. This model is famous for forgetting about world events from 1989, but also seems to have a few issues when faced with concrete coding tasks. It is a reasoning model, meaning it “thinks” before acting, which should lead to improved accuracy. In practice, it regularly gets stuck indefinitely searching where it should start and jumping from one problem to the other in a loop. This can happen even on simple problems, and made it unusable for me. mistral:7b Mistral is the French alternative to American and Chinese models. I have already talked about their 7b model on this blog. It is worth noting that they have kept updating their models, and it should now be much more accurate than two years ago. Mistral is not a reasoning model, so it will jump straight to answering. This is very good if you’re working with tasks where speed and low compute use are a priority. Sadly, the accuracy doesn’t seem good enough for coding. Even on simple tasks, it will hallucinate functions or randomly delete parts of the code I didn’t want to touch. qwen3:8b Another model from China, qwen3 was created by the folks at Alibaba. It also claims impressive benchmark results, and can work as both a reasoning or non-thinking model. It was made with modern AI tooling in mind, by supporting MCPs and a framework for agentic development. This model actually seems to work as expected, providing somewhat accurate code output while not hanging in the reasoning part. Since it runs decently on my local setup, I decided to stick to that model for now. Setting up a local API with Ollama Ollama is now the default way to download and run local LLMs. It can be simply installed by downloading it from their website. Once installed, it works like Docker for models, by giving us access to commands like pull, run, or rm. Ollama will expose an API on localhost which can be used by other programs. For example, you can use it from your Python programs through ollama-python. Pair programming with aider The next piece of software I installed is aider. I assume it’s pronounced like the French word, but I could not confirm that. Aider describes itself as a “pair programming” application. Its main job is to pass context to the model, let it write the output to files, run linters, and commit the changes. Getting started It can be installed using the official Python package or via Homebrew if you use Mac. Once it is installed, just navigate to your code repository and launch it: export OLLAMA_API_BASE=http://127.0.0.1:11434 aider --model ollama_chat/qwen3:8b The CLI should automatically create some configuration files and add them to the repo’s .gitignore. Usage Aider isn’t meant to be left alone in complete autonomy. You’ll have to guide the AI through the process of making changes to your repository. To start, use the /add command to add files you want to focus on. Those files will be passed entirely to the model’s context and the model will be able to write in them. You can then ask questions using the /ask command. If you want to generate code, a good strategy can be to starting by requesting a plan of actions. When you want it to actually write to the files, you can prompt it using the /code command. This is also the default mode. There’s no absolute guarantee that it will follow a plan if you agreed on one previously, but it is still a good idea to have one. The /architect command seems to automatically ask for a plan, accept it, and write the code. The specificity of this command is that it lets you use different models to plan and write the changes. Refactoring I tried coding with aider in a few situations to see how it performs in practice. First, I tried making it do a simple refactoring on Itako, which is a project of average complexity. When pointed to the exact part of code where the issues happened, and explained explicitly what to do, the model managed to change the target struct according to the instructions. It did unexpectedly change a function that was outside the scope of what I asked, but this was easy to spot. On paper, this looks like a success. In practice, the time spent crafting a prompt, waiting for the AI to run and fixing the small issue that came up immensely exceeds the 10 minutes it would have taken me to edit the file myself. I don’t think coding that way would lead me to a massive performance improvement for now. Greenfield project For a second scenario, I wanted to see how it would perform on a brand-new project. I quickly set up a Python virtual environment, and asked aider to work with me at building a simple project. We would be opening a file containing Japanese text, parsing it with fugashi, and counting the words. To my surprise, this was a disaster. All I got was a bunch on hallucination riddled python that wouldn’t run under any circumstances. It may be that the lack of context actually made it harder for the model to generate code. Troubleshooting Finally, I went back to Itako, and decided to check how it would perform on common troubleshooting tasks. I introduced a few bugs to my code and gathered some error messages. I then proceeded to simply give aider the files mentioned by the error message and just use /ask to have it explain the errors to me, without requiring it to implement the code. This part did work very well. If I compare it with Googling unknown error messages, I think this can cut the time spent on the issue by half This is not just because Google is getting worse every day, but the model having access to the actual code does give it a massive advantage. I do think this setup is something I can use instead of the occasional frustration of scrolling through StackOverflow threads when something unexpected breaks. What about the Qwen CLI? With everyone jumping on the trend of CLI tools for LLMs, the Qwen team released its own Qwen Code. It can be installed using npm, and connects to a local model if configured like this: export OPENAI_API_KEY="ollama" export OPENAI_BASE_URL="http://localhost:11434/v1/" export OPENAI_MODEL="qwen3:8b" Compared to aider, it aims at being fully autonomous. For example, it will search your repository using grep. However, I didn’t manage to get it to successfully write any code. The tool seems optimized for larger, online models, with context sizes up to 1M tokens. Our local qwen3 context only has a 40k tokens context size, which can get overwhelmed very quickly when browsing entire code repositories. Even when I didn’t run out of context, the tool mysteriously failed when trying to write files. It insists it can only write to absolute paths, which the model doesn’t seem to agree with providing. I did not investigate the issue further.

yesterday 5 votes
Building a container orchestrator

Kubernetes is not exactly the most fun piece of technology around. Learning it isn’t easy, and learning the surrounding ecosystem is even harder. Even those who have managed to tame it are still afraid of getting paged by an ETCD cluster corruption, a Kubelet certificate expiration, or the DNS breaking down (and somehow, it’s always the DNS). Samuel Sianipar If you’re like me, the thought of making your own orchestrator has crossed your mind a few times. The result would, of course, be a magical piece of technology that is both simple to learn and wouldn’t break down every weekend. Sadly, the task seems daunting. Kubernetes is a multi-million lines of code project which has been worked on for more than a decade. The good thing is someone wrote a book that can serve as a good starting point to explore the idea of building our own container orchestrator. This book is named “Build an Orchestrator in Go”, written by Tim Boring, published by Manning. The tasks The basic unit of our container orchestrator is called a “task”. A task represents a single container. It contains configuration data, like the container’s name, image and exposed ports. Most importantly, it indicates the container state, and so acts as a state machine. The state of a task can be Pending, Scheduled, Running, Completed or Failed. Each task will need to interact with a container runtime, through a client. In the book, we use Docker (aka Moby). The client will get its configuration from the task and then proceed to pull the image, create the container and start it. When it is time to finish the task, it will stop the container and remove it. The workers Above the task, we have workers. Each machine in the cluster runs a worker. Workers expose an API through which they receive commands. Those commands are added to a queue to be processed asynchronously. When the queue gets processed, the worker will start or stop tasks using the container client. In addition to exposing the ability to start and stop tasks, the worker must be able to list all the tasks running on it. This demands keeping a task database in the worker’s memory and updating it every time a task change’s state. The worker also needs to be able to provide information about its resources, like the available CPU and memory. The book suggests reading the /proc Linux file system using goprocinfo, but since I use a Mac, I used gopsutil. The manager On top of our cluster of workers, we have the manager. The manager also exposes an API, which allows us to start, stop, and list tasks on the cluster. Every time we want to create a new task, the manager will call a scheduler component. The scheduler has to list the workers that can accept more tasks, assign them a score by suitability and return the best one. When this is done, the manager will send the work to be done using the worker’s API. In the book, the author also suggests that the manager component should keep track of every tasks state by performing regular health checks. Health checks typically consist of querying an HTTP endpoint (i.e. /ready) and checking if it returns 200. In case a health check fails, the manager asks the worker to restart the task. I’m not sure if I agree with this idea. This could lead to the manager and worker having differing opinions about a task state. It will also cause scaling issues: the manager workload will have to grow linearly as we add tasks, and not just when we add workers. As far as I know, in Kubernetes, Kubelet (the equivalent of the worker here) is responsible for performing health checks. The CLI The last part of the project is to create a CLI to make sure our new orchestrator can be used without having to resort to firing up curl. The CLI needs to implement the following features: start a worker start a manager run a task in the cluster stop a task get the task status get the worker node status Using cobra makes this part fairly straightforward. It lets you create very modern feeling command-line apps, with properly formatted help commands and easy argument parsing. Once this is done, we almost have a fully functional orchestrator. We just need to add authentication. And maybe some kind of DaemonSet implementation would be nice. And a way to handle mounting volumes…

4 weeks ago 19 votes
Discord considered harmful

In the past few years, social media use has gained a bad reputation. More or less everyone is now aware that TikTok is ruining your attention span, and Twitter is radicalizing you into extreme ideologies. But, despite its enormous popularity amongst technology enthusiasts, there’s not a lot of attention given to Discord. I personally have been using Discord so much for so long that the majority of my social circle is made of people I met through the platform. I even spent two years of my life helping run the infrastructure behind the most popular Bot available on Discord. In this article, I will try to give my perspective on Discord, why I think it is harmful, and what can we do about it. appshunter.io A tale of two book clubs To explain my point of view about Discord, I will compare the experience between joining a real-life book-club, and one that communicates exclusively through Discord. This example is about books, but the same issues would apply if it was a community talking about investing, knitting, or collecting stamps. As Marshall McLuhan showed last century, examining media should be done independently of their content. In the first scenario, we have Bob. Bob enjoys reading books, which is generally a solitary hobby. To break this solitude, Bob decides to join a book club. This book club reunites twice a month in a library where they talk about a new book each time. In the second scenario, we have Alice. Alice also likes books. Alice also wants to meet fellow book lovers. Being a nerd, Alice decides to join a Discord server. This server does not have fixed meeting times. Most users simply use the text channels to talk about what they are reading anytime during the day. Crumbs of Belongingness In Bob’s book club, a session typically lasts an hour. First, the librarian takes some time to welcome everyone and introduce newcomers. After, that each club member talks about the book they were expected to read. They can talk about what they liked and disliked, how the book made them feel, and the chapters they found particularly noteworthy. Once each member had the time to talk about the book, they vote on the book they are going to read for the next date. After the session is concluded, some members move to the nearest coffeehouse to keep talking. During this session of one hour, Bob spent around one hour socializing. The need for belongingness that drove Bob to join this book club is fully met. On Alice’s side, the server is running 24/7. When she opens the app, even if there are sometimes more than 4000 members of her virtual book club online, most of the time, nobody is talking. If she was to spend an entire hour staring at the server she might witness a dozen or so messages. Those messages may be part of small conversations in which Alice can take part. Sadly, most of the time they will be simple uploads of memes, conversations about books she hasn’t read, or messages that do not convey enough meaning to start a conversation. In one hour of constant Discord use, Alice’s need for socializing has not been met. Susan Q Yin The shop is closed Even if Bob’s library is open every day, the book club is only open for a total of two hours a month. It is enough for Bob. Since the book club fulfills his need, he doesn’t want it to be around for longer. He has not even entertained the thought of joining a second book club, because too many meetings would be overwhelming. For Alice, Discord is always available. No matter if she is at home or away, it is always somewhere in her phone or taskbar. At any moment of the day, she might notice a red circle above the icon. It tells her there are unread messages on Discord. When she notices that, she instinctively stops her current task and opens the app to spend a few minutes checking her messages. Most of the time those messages do not lead to a meaningful conversation. Reading a few messages isn’t enough to meet her need for socialization. So, after having scrolled through the messages, she goes back to waiting for the next notification. Each time she interrupts her current task to check Discord, getting back into the flow can take several minutes or not happen at all. This can easily happen dozens of times a day and cost Alice hundreds of hours each month. Book hopping When Bob gets home, the club only requires him to read the next book. He may also choose to read two books at the same time, one for the book club and one from his personal backlog. But, if he were to keep his efforts to a strict minimum, he would still have things to talk about in the next session. Alice wants to be able to talk with other users about the books they are reading. So she starts reading the books that are trending and get mentionned often. The issue is, Discord’s conversation are instantaneous, and instantaneity compresses time. A book isn’t going to stay popular and relevant for two whole weeks, if it manages to be the thing people talk about for two whole days, it’s already great. Alice might try to purchase and read two to three books a week to keep up with the server rythm. Even if books are not terribly expensive, this can turn a 20 $/month hobby into a 200 $/month hobby. In addition to that, if reading a book takes Alice on average 10 hours, reading 3 books a week would be like adding a part-time job to her schedule. All this, while being constantly interrupted by the need to check if new conversations have been posted to the server. visnu deva Quitting Discord If you are in Alice’s situation, the solution is quite simple: use Discord less, ideally not at all. On my side, I’ve left every server that is not relevant to my current work. I blocked discord.com from the DNS of my coding computer (using NextDNS) and uninstalled the app from my phone. This makes the platform only usable as a direct messaging app, exclusively from my gaming device, which I cannot carry with me. I think many people realize the addictive nature of Discord, yet keep using the application all the time. One common objection to quitting the platform, is that there’s a need for an alternative: maybe we should go back to forums, or IRC, or use Matrix, etc… I don’t think any alternative internet chat platform can solve the problem. The real problem is that we want to be able to talk to people without leaving home, at any time, without any inconvenience. But what we should do is exactly that, leave home and join a real book club, one that is not open 24/7, and one where the members take the time to listen to each other. In the software community, we have also been convinced that every one of our projects needs to be on Discord. Every game needs a server, open-source projects offer support on Discord, and a bunch of AI startups even use it as their main user interface. I even made a server for Dice’n Goblins. I don’t think it’s really that useful. I’m not even sure it’s that convenient. Popular games are not popular because they have big servers, they have big servers because they are popular. Successful open-source projects often don’t even have a server.

a month ago 21 votes
Thoughts on releasing our first indie game

Two weeks ago we released Dice’n Goblins, our first game on Steam. This project allowed me to discover and learn a lot of new things about game development and the industry. I will use this blog post to write down what I consider to be the most important lessons from the months spent working on this. The development started around 2 years ago when Daphnée started prototyping a dungeon crawler featuring a goblin protagonist. After a few iterations, the game combat started featuring dice, and then those dice could be used to make combos. In May 2024, the game was baptized Dice’n Goblins, and a Steam page was created featuring some early gameplay screenshots and footage. I joined the project full-time around this period. Almost one year later, after amassing more than 8000 wishlists, the game finally released on Steam on April 4th, 2025. It was received positively by the gaming press, with great reviews from PCGamer and LadiesGamers. It now sits at 92% positive reviews from players on Steam. Building RPGs isn’t easy As you can see from the above timeline, building this game took almost two years and two programmers. This is actually not that long if you consider that other indie RPGs have taken more than 6 years to come out. The main issue with the genre is that you need to create a believable world. In practice, this requires programming many different systems that will interact together to give the impression of a cohesive universe. Every time you add a new system, you need to think about how it will fit all the existing game features. For example, players typically expect an RPG to have a shop system. Of course, this means designing a shop non-player character (or building) and creating a UI that is displayed when you interact with it. But this also means thinking through a lot of other systems: combat needs to be changed to reward the player with gold, every item needs a price tag, chests should sometimes reward the player with gold, etc… Adding too many systems can quickly get into scope creep territory, and make the development exponentially longer. But you can only get away with removing so much until your game stops being an RPG. Making a game without a shop might be acceptable, but the experience still needs to have more features than “walking around and fighting monsters” to feel complete. RPGs are also, by definition, narrative experiences. While some games have managed to get away with procedurally generating 90% of the content, in general, you’ll need to get your hands dirty, write a story, and design a bunch of maps. Creating enough content for a game to fit 12h+ without having the player go through repetitive grind will by itself take a lot of time. Having said all that, I definitely wouldn’t do any other kind of games than RPGs, because this is what I enjoy playing. I don’t think I would be able to nail what makes other genres fun if I don’t play them enough to understand what separates the good from the mediocre. Marketing isn’t that complicated Everyone in the game dev community knows that there are way too many games releasing on Steam. To stand out amongst the 50+ games coming out every day, it’s important not only to have a finished product but also to plan a marketing campaign well in advance. For most people coming from a software engineering background, like me, this can feel extremely daunting. Our education and jobs do not prepare us well for this kind of task. In practice, it’s not that complicated. If your brain is able to provision a Kubernetes cluster, then you are most definitely capable of running a marketing campaign. Like anything else, it’s a skill that you can learn over time by practicing it, and iteratively improving your methods. During the 8 months following the Steam page release, we tried basically everything you can think of as a way to promote the game. Every time something was having a positive impact, we would do it more, and we quickly stopped things with low impact. The most important thing to keep in mind is your target audience. If you know who wants the game of games you’re making, it is very easy to find where they hang out and talk to them. This is however not an easy question to answer for every game. For a long while, we were not sure who would like Dice’n Goblins. Is it people who like Etrian Odyssey? Fans of Dicey Dungeons? Nostalgic players of Paper Mario? For us, the answer was mostly #1, with a bit of #3. Once we figured out what was our target audience, how to communicate with them, and most importantly, had a game that was visually appealing enough, marketing became very straightforward. This is why we really struggled to get our first 1000 wishlists, but getting the last 5000 was actually not that complicated. Publishers aren’t magic At some point, balancing the workload of actually building the game and figuring out how to market it felt too much for a two-person team. We therefore did what many indie studios do, and decided to work with a publisher. We worked with Rogue Duck Interactive, who previously published Dice & Fold, a fairly successful dice roguelike. Without getting too much into details, it didn’t work out as planned and we decided, by mutual agreement, to go back to self-publishing Dice’n Goblins. The issue simply came from the audience question mentioned earlier. Even though Dice & Fold and Dice’n Goblins share some similarities, they target a different audience, which requires a completely different approach to marketing. The lesson learned is that when picking a publisher, the most important thing you can do is to check that their current game catalog really matches the idea you have of your own game. If you’re building a fast-paced FPS, a publisher that only has experience with cozy simulation games will not be able to help you efficiently. In our situation, a publisher with experience in roguelikes and casual strategy games wasn’t a good fit for an RPG. In addition to that, I don’t think the idea of using a publisher to remove marketing toil and focus on making the game is that much of a good idea in the long term. While it definitely helps to remove the pressure from handling social media accounts and ad campaigns, new effort will be required in communicating and negotiating with the publishing team. In the end, the difference between the work saved and the work gained might not have been worth selling a chunk of your game. Conclusion After all this was said and done, one big question I haven’t answered is: would I do it again? The answer is definitely yes. Not only building this game was an extremely satisfying endeavor, but so much has been learned and built while doing it, it would be a shame not to go ahead and do a second one.

3 months ago 38 votes
Writing GDScript with Neovim

Neovim is by far my favorite text editor. The clutter-free interface and keyboard-only navigation are what keep me productive in my daily programming. In an earlier post, I explained how I configure it into a minimalist development environment. Today, I will show you how to use it with Godot and GDScript. Configure Godot First, we need to tell Godot to use nvim as a text editor instead of the built-in one. Open Godot, and head to Editor Settings > General > Text Editor > External. There, you will need to tick the box Use external editor, indicate your Neovim installation path, and use --server /tmp/godothost --remote-send "<C-\><C-N>:n {file}<CR>{line}G{col}|" as execution flags. While in the settings, head to Network > Language Server and note down the remote port Godot is using. By default, it should be 6005. We will need that value later. Connecting to Godot with vim-godot Neovim will be able to access Godot features by using a plugin called vim-godot. We will need to edit the nvim configuration file to install plugins and configure Neovim. On Mac and Linux, it is located at ~/.config/nvim/init.vim I use vim-plug to manage my plugins, so I can just add it to my configuration like this: call plug#begin('~/.vim/plugged') " ... Plug 'habamax/vim-godot' " ... call plug#end() Once the configuration file is modified and saved, use the :PlugInstall command to install it. You’ll also need to indicate Godot’s executable path. Add this line to your init.vim: let g:godot_executable = '/Applications/Godot.app/Contents/MacOS/Godot' For vim-godot to communicate with the Godot editor, it will need to listen to the /tmp/godothost file we configured in the editor previously. To do that, simply launch nvim with the flag --listen /tmp/godothost. To save you some precious keypress, I suggest creating a new alias in your bashrc/zshrc like this: alias gvim="nvim --listen /tmp/godothost" Getting autocompletion with coc.nvim Godot ships with a language server. It means the Godot editor can provide autocompletion, syntax highlighting, and advanced navigation to external editors like nvim. While Neovim now has built-in support for the language server protocol, I’ve used the plugin coc.nvim to obtain these functionalities for years and see no reason to change. You can also install it with vim-plug by adding the following line to your plugin list: Plug 'neoclide/coc.nvim', {'branch':'release'} Run :PlugInstall again to install it. You’ll need to indicate the Godot language server address and port using the command :CocConfig. It should open Coc’s configuration file, which is a JSON file normally located at ~/.config/nvim/coc-settings.json. In this file enter the following data, and make sure the port number matches the one located in your editor: { "languageserver": { "godot": { "host": "127.0.0.1", "filetypes": ["gdscript"], "port": 6005 } } } I recommend adding Coc’s example configuration to your init.vim file. You can find it on GitHub. It will provide you with a lot of useful shortcuts, such as using gd to go to a function definition and gr to list its references. Debugging using nvim-dap If you want to use the debugger from inside Neovim, you’ll need to install another plugin called nvim-dap. Add the following to your plugins list: Plug 'mfussenegger/nvim-dap' The plugin authors suggest configuring it using Lua, so let’s do that by adding the following in your init.vim: lua <<EOF local dap = require("dap") dap.adapters.godot = { type = "server", host = "127.0.0.1", port = 6006, } dap.configurations.gdscript = { { type = "godot", request = "launch", name = "Launch scene", project = "${workspaceFolder}", launch_scene = true, }, } vim.api.nvim_create_user_command("Breakpoint", "lua require'dap'.toggle_breakpoint()", {}) vim.api.nvim_create_user_command("Continue", "lua require'dap'.continue()", {}) vim.api.nvim_create_user_command("StepOver", "lua require'dap'.step_over()", {}) vim.api.nvim_create_user_command("StepInto", "lua require'dap'.step_into()", {}) vim.api.nvim_create_user_command("REPL", "lua require'dap'.repl.open()", {}) EOF This will connect to the language server (here on port 6005), and allow you to pilot the debugger using the following commands: :Breakpoint to create (or remove) a breakpoint :Continue to launch the game or run until the next breakpoint :StepOver to step over a line :StepInto to step inside a function definition :REPL to launch a REPL (useful if you want to examine values) Conclusion I hope you’ll have a great time developing Godot games with Neovim. If it helps you, you can check out my entire init.vim file on GitHub gist.

9 months ago 81 votes

More in programming

Our $100M Series B

We don’t want to bury the lede: we have raised a $100M Series B, led by a new strategic partner in USIT with participation from all existing Oxide investors. To put that number in perspective: over the nearly six year lifetime of the company, we have raised $89M; our $100M Series B more than doubles our total capital raised to date — and positions us to make Oxide the generational company that we have always aspired it to be. If this aspiration seems heady now, it seemed absolutely outlandish when we were first raising venture capital in 2019. Our thesis was that cloud computing was the future of all computing; that running on-premises would remain (or become!) strategically important for many; that the entire stack — hardware and software — needed to be rethought from first principles to serve this market; and that a large, durable, public company could be built by whomever pulled it off. This scope wasn’t immediately clear to all potential investors, some of whom seemed to latch on to one aspect or another without understanding the whole. Their objections were revealing: "We know you can build this," began more than one venture capitalist (at which we bit our tongue; were we not properly explaining what we intended to build?!), "but we don’t think that there is a market." Entrepreneurs must become accustomed to rejection, but this flavor was particularly frustrating because it was exactly backwards: we felt that there was in fact substantial technical risk in the enormity of the task we put before ourselves — but we also knew that if we could build it (a huge if!) there was a huge market, desperate for cloud computing on-premises. Fortunately, in Eclipse Ventures we found investors who saw what we saw: that the most important products come when we co-design hardware and software together, and that the on-premises market was sick of being told that they either don’t exist or that they don’t deserve modernity. These bold investors — like the customers we sought to serve — had been waiting for this company to come along; we raised seed capital, and started building. And build it we did, making good on our initial technical vision: We did our own board designs, allowing for essential system foundation like a true hardware root-of-trust and end-to-end power observability. We did our own microcontroller operating system, and used it to replace the traditional BMC. We did our own platform enablement software, eliminating the traditional UEFI BIOS and its accompanying flotilla of vulnerabilities. We did our own host hypervisor, assuring an integrated and seamless user experience — and eliminating the need for a third-party hypervisor and its concomitant rapacious software licensing. We did our own switch — and our own switch runtime — eliminating entire universes of integration complexity and operational nightmares. We did our own integrated storage service, allowing the rack-scale system to have reliable, available, durable, elastic instance storage without necessitating a dependency on a third party. We did our own control plane, a sophisticated distributed system building on the foundation of our hardware and software components to deliver the API-driven services that modernity demands: elastic compute, virtual networking, and virtual storage. While these technological components are each very important (and each is in service to specific customer problems when deploying infrastructure on-premises), the objective is the product, not its parts. The journey to a product was long, but we ticked off the milestones. We got the boards brought up. We got the switch transiting packets. We got the control plane working. We got the rack manufactured. We passed FCC compliance. And finally, two years ago, we shipped our first system! Shortly thereafter, more milestones of the variety you can only get after shipping: our first update of the software in the field; our first update-delivered performance improvements; our first customer-requested features added as part of an update. Later that year, we hit general commercial availability, and things started accelerating. We had more customers — and our first multi-rack customer. We had customers go on the record about why they had selected Oxide — and customers describing the wins that they had seen deploying Oxide. Customers starting landing faster now: enterprise sales cycles are infamously long, but we were finding that we were going from first conversations to a delivered product surprisingly quickly. The quickening pace always seemed to be due in some way to our transparency: new customers were listeners to our podcast, or they had read our RFDs, or they had perused our documentation, or they had looked at the source code itself. With growing customer enthusiasm, we were increasingly getting questions about what it would look like to buy a large number of Oxide racks. Could we manufacture them? Could we support them? Could we make them easy to operate together? Into this excitement, a new potential investor, USIT, got to know us. They asked terrific questions, and we found a shared disposition towards building lasting value and doing it the right way. We learned more about them, too, and especially USIT’s founder, Thomas Tull. The more we each learned about the other, the more there was to like. And importantly, USIT had the vision for us that we had for ourselves: that there was a big, important market here — and that it was uniquely served by Oxide. We are elated to announce this new, exciting phase of the company. It’s not necessarily in our nature to celebrate fundraising, but this is a big milestone, because it will allow us to address our customers' most pressing questions around scale (manufacturing scale, system scale, operations scale) and roadmap scope. We have always believed in our mission, but this raise gives us a new sense of confidence when we say it: we’re going to kick butt, have fun, not cheat (of course!), love our customers — and change computing forever.

7 hours ago 4 votes
2000 words about arrays and tables

I'm way too discombobulated from getting next month's release of Logic for Programmers ready, so I'm pulling a idea from the slush pile. Basically I wanted to come up with a mental model of arrays as a concept that explained APL-style multidimensional arrays and tables but also why there weren't multitables. So, arrays. In all languages they are basically the same: they map a sequence of numbers (I'll use 1..N)1 to homogeneous values (values of a single type). This is in contrast to the other two foundational types, associative arrays (which map an arbitrary type to homogeneous values) and structs (which map a fixed set of keys to heterogeneous values). Arrays appear in PLs earlier than the other two, possibly because they have the simplest implementation and the most obvious application to scientific computing. The OG FORTRAN had arrays. I'm interested in two structural extensions to arrays. The first, found in languages like nushell and frameworks like Pandas, is the table. Tables have string keys like a struct and indexes like an array. Each row is a struct, so you can get "all values in this column" or "all values for this row". They're heavily used in databases and data science. The other extension is the N-dimensional array, mostly seen in APLs like Dyalog and J. Think of this like arrays-of-arrays(-of-arrays), except all arrays at the same depth have the same length. So [[1,2,3],[4]] is not a 2D array, but [[1,2,3],[4,5,6]] is. This means that N-arrays can be queried on any axis. ]x =: i. 3 3 0 1 2 3 4 5 6 7 8 0 { x NB. first row 0 1 2 0 {"1 x NB. first column 0 3 6 So, I've had some ideas on a conceptual model of arrays that explains all of these variations and possibly predicts new variations. I wrote up my notes and did the bare minimum of editing and polishing. Somehow it ended up being 2000 words. 1-dimensional arrays A one-dimensional array is a function over 1..N for some N. To be clear this is math functions, not programming functions. Programming functions take values of a type and perform computations on them. Math functions take values of a fixed set and return values of another set. So the array [a, b, c, d] can be represented by the function (1 -> a ++ 2 -> b ++ 3 -> c ++ 4 -> d). Let's write the set of all four element character arrays as 1..4 -> char. 1..4 is the function's domain. The set of all character arrays is the empty array + the functions with domain 1..1 + the functions with domain 1..2 + ... Let's call this set Array[Char]. Our compilers can enforce that a type belongs to Array[Char], but some operations care about the more specific type, like matrix multiplication. This is either checked with the runtime type or, in exotic enough languages, with static dependent types. (This is actually how TLA+ does things: the basic collection types are functions and sets, and a function with domain 1..N is a sequence.) 2-dimensional arrays Now take the 3x4 matrix i. 3 4 0 1 2 3 4 5 6 7 8 9 10 11 There are two equally valid ways to represent the array function: A function that takes a row and a column and returns the value at that index, so it would look like f(r: 1..3, c: 1..4) -> Int. A function that takes a row and returns that column as an array, aka another function: f(r: 1..3) -> g(c: 1..4) -> Int.2 Man, (2) looks a lot like currying! In Haskell, functions can only have one parameter. If you write (+) 6 10, (+) 6 first returns a new function f y = y + 6, and then applies f 10 to get 16. So (+) has the type signature Int -> Int -> Int: it's a function that takes an Int and returns a function of type Int -> Int.3 Similarly, our 2D array can be represented as an array function that returns array functions: it has type 1..3 -> 1..4 -> Int, meaning it takes a row index and returns 1..4 -> Int, aka a single array. (This differs from conventional array-of-arrays because it forces all of the subarrays to have the same domain, aka the same length. If we wanted to permit ragged arrays, we would instead have the type 1..3 -> Array[Int].) Why is this useful? A couple of reasons. First of all, we can apply function transformations to arrays, like "combinators". For example, we can flip any function of type a -> b -> c into a function of type b -> a -> c. So given a function that takes rows and returns columns, we can produce one that takes columns and returns rows. That's just a matrix transposition! Second, we can extend this to any number of dimensions: a three-dimensional array is one with type 1..M -> 1..N -> 1..O -> V. We can still use function transformations to rearrange the array along any ordering of axes. Speaking of dimensions: What are dimensions, anyway Okay, so now imagine we have a Row × Col grid of pixels, where each pixel is a struct of type Pixel(R: int, G: int, B: int). So the array is Row -> Col -> Pixel But we can also represent the Pixel struct with a function: Pixel(R: 0, G: 0, B: 255) is the function where f(R) = 0, f(G) = 0, f(B) = 255, making it a function of type {R, G, B} -> Int. So the array is actually the function Row -> Col -> {R, G, B} -> Int And then we can rearrange the parameters of the function like this: {R, G, B} -> Row -> Col -> Int Even though the set {R, G, B} is not of form 1..N, this clearly has a real meaning: f[R] is the function mapping each coordinate to that coordinate's red value. What about Row -> {R, G, B} -> Col -> Int? That's for each row, the 3 × Col array mapping each color to that row's intensities. Really any finite set can be a "dimension". Recording the monitor over a span of time? Frame -> Row -> Col -> Color -> Int. Recording a bunch of computers over some time? Computer -> Frame -> Row …. This is pretty common in constraint satisfaction! Like if you're conference trying to assign talks to talk slots, your array might be type (Day, Time, Room) -> Talk, where Day/Time/Room are enumerations. An implementation constraint is that most programming languages only allow integer indexes, so we have to replace Rooms and Colors with numerical enumerations over the set. As long as the set is finite, this is always possible, and for struct-functions, we can always choose the indexing on the lexicographic ordering of the keys. But we lose type safety. Why tables are different One more example: Day -> Hour -> Airport(name: str, flights: int, revenue: USD). Can we turn the struct into a dimension like before? In this case, no. We were able to make Color an axis because we could turn Pixel into a Color -> Int function, and we could only do that because all of the fields of the struct had the same type. This time, the fields are different types. So we can't convert {name, flights, revenue} into an axis. 4 One thing we can do is convert it to three separate functions: airport: Day -> Hour -> Str flights: Day -> Hour -> Int revenue: Day -> Hour -> USD But we want to keep all of the data in one place. That's where tables come in: an array-of-structs is isomorphic to a struct-of-arrays: AirportColumns( airport: Day -> Hour -> Str, flights: Day -> Hour -> Int, revenue: Day -> Hour -> USD, ) The table is a sort of both representations simultaneously. If this was a pandas dataframe, df["airport"] would get the airport column, while df.loc[day1] would get the first day's data. I don't think many table implementations support more than one axis dimension but there's no reason they couldn't. These are also possible transforms: Hour -> NamesAreHard( airport: Day -> Str, flights: Day -> Int, revenue: Day -> USD, ) Day -> Whatever( airport: Hour -> Str, flights: Hour -> Int, revenue: Hour -> USD, ) In my mental model, the heterogeneous struct acts as a "block" in the array. We can't remove it, we can only push an index into the fields or pull a shared column out. But there's no way to convert a heterogeneous table into an array. Actually there is a terrible way Most languages have unions or product types that let us say "this is a string OR integer". So we can make our airport data Day -> Hour -> AirportKey -> Int | Str | USD. Heck, might as well just say it's Day -> Hour -> AirportKey -> Any. But would anybody really be mad enough to use that in practice? Oh wait J does exactly that. J has an opaque datatype called a "box". A "table" is a function Dim1 -> Dim2 -> Box. You can see some examples of what that looks like here Misc Thoughts and Questions The heterogeneity barrier seems like it explains why we don't see multiple axes of table columns, while we do see multiple axes of array dimensions. But is that actually why? Is there a system out there that does have multiple columnar axes? The array x = [[a, b, a], [b, b, b]] has type 1..2 -> 1..3 -> {a, b}. Can we rearrange it to 1..2 -> {a, b} -> 1..3? No. But we can rearrange it to 1..2 -> {a, b} -> PowerSet(1..3), which maps rows and characters to columns with that character. [(a -> {1, 3} ++ b -> {2}), (a -> {} ++ b -> {1, 2, 3}]. We can also transform Row -> PowerSet(Col) into Row -> Col -> Bool, aka a boolean matrix. This makes sense to me as both forms are means of representing directed graphs. Are other function combinators useful for thinking about arrays? Does this model cover pivot tables? Can we extend it to relational data with multiple tables? Systems Distributed Talk (will be) Online The premier will be August 6 at 12 CST, here! I'll be there to answer questions / mock my own performance / generally make a fool of myself. Sacrilege! But it turns out in this context, it's easier to use 1-indexing than 0-indexing. In the years since I wrote that article I've settled on "each indexing choice matches different kinds of mathematical work", so mathematicians and computer scientists are best served by being able to choose their index. But software engineers need consistency, and 0-indexing is overall a net better consistency pick. ↩ This is right-associative: a -> b -> c means a -> (b -> c), not (a -> b) -> c. (1..3 -> 1..4) -> Int would be the associative array that maps length-3 arrays to integers. ↩ Technically it has type Num a => a -> a -> a, since (+) works on floats too. ↩ Notice that if each Airport had a unique name, we could pull it out into AirportName -> Airport(flights, revenue), but we still are stuck with two different values. ↩

7 hours ago 2 votes
React Server Components with Vite and React-Router (tip)

Create a small example app and send payloads from the server to the client using RSC's

an hour ago 2 votes
Playing with open source LLMs

Every 6 months or so, I decide to leave my cave and check out what the cool kids are doing with AI. Apparently the latest trend is to use fancy command line tools to write code using LLMs. This is a very nice change, since it suddenly makes AI compatible with my allergy to getting out of the terminal. The most popular of these tools seems to be Claude Code. It promises to be able to build in total autonomy, being able to use search code, write code, run tests, lint, and commit the changes. While this sounds great on paper, I’m not keen on getting locked into vendor tools from an unprofitable company. At some point, they will either need to raise their prices, enshittify their product, or most likely do both. So I went looking for what the free and open source alternatives are. Picking a model There’s a large amount of open source large language models on the market, with new ones getting released all the time. However, they are not all ready to be used locally in coding tasks, so I had to try a bunch of them before settling on one. deepseek-r1:8b Deepseek is the most popular open source model right now. It was created by the eponymous Chinese company. It made the news by beating numerous benchmarks while being trained on a budget that is probably lower than the compensation of some OpenAI workers. The 8b variant only weights 5.2 GB and runs decently on limited hardware, like my three years old Mac. This model is famous for forgetting about world events from 1989, but also seems to have a few issues when faced with concrete coding tasks. It is a reasoning model, meaning it “thinks” before acting, which should lead to improved accuracy. In practice, it regularly gets stuck indefinitely searching where it should start and jumping from one problem to the other in a loop. This can happen even on simple problems, and made it unusable for me. mistral:7b Mistral is the French alternative to American and Chinese models. I have already talked about their 7b model on this blog. It is worth noting that they have kept updating their models, and it should now be much more accurate than two years ago. Mistral is not a reasoning model, so it will jump straight to answering. This is very good if you’re working with tasks where speed and low compute use are a priority. Sadly, the accuracy doesn’t seem good enough for coding. Even on simple tasks, it will hallucinate functions or randomly delete parts of the code I didn’t want to touch. qwen3:8b Another model from China, qwen3 was created by the folks at Alibaba. It also claims impressive benchmark results, and can work as both a reasoning or non-thinking model. It was made with modern AI tooling in mind, by supporting MCPs and a framework for agentic development. This model actually seems to work as expected, providing somewhat accurate code output while not hanging in the reasoning part. Since it runs decently on my local setup, I decided to stick to that model for now. Setting up a local API with Ollama Ollama is now the default way to download and run local LLMs. It can be simply installed by downloading it from their website. Once installed, it works like Docker for models, by giving us access to commands like pull, run, or rm. Ollama will expose an API on localhost which can be used by other programs. For example, you can use it from your Python programs through ollama-python. Pair programming with aider The next piece of software I installed is aider. I assume it’s pronounced like the French word, but I could not confirm that. Aider describes itself as a “pair programming” application. Its main job is to pass context to the model, let it write the output to files, run linters, and commit the changes. Getting started It can be installed using the official Python package or via Homebrew if you use Mac. Once it is installed, just navigate to your code repository and launch it: export OLLAMA_API_BASE=http://127.0.0.1:11434 aider --model ollama_chat/qwen3:8b The CLI should automatically create some configuration files and add them to the repo’s .gitignore. Usage Aider isn’t meant to be left alone in complete autonomy. You’ll have to guide the AI through the process of making changes to your repository. To start, use the /add command to add files you want to focus on. Those files will be passed entirely to the model’s context and the model will be able to write in them. You can then ask questions using the /ask command. If you want to generate code, a good strategy can be to starting by requesting a plan of actions. When you want it to actually write to the files, you can prompt it using the /code command. This is also the default mode. There’s no absolute guarantee that it will follow a plan if you agreed on one previously, but it is still a good idea to have one. The /architect command seems to automatically ask for a plan, accept it, and write the code. The specificity of this command is that it lets you use different models to plan and write the changes. Refactoring I tried coding with aider in a few situations to see how it performs in practice. First, I tried making it do a simple refactoring on Itako, which is a project of average complexity. When pointed to the exact part of code where the issues happened, and explained explicitly what to do, the model managed to change the target struct according to the instructions. It did unexpectedly change a function that was outside the scope of what I asked, but this was easy to spot. On paper, this looks like a success. In practice, the time spent crafting a prompt, waiting for the AI to run and fixing the small issue that came up immensely exceeds the 10 minutes it would have taken me to edit the file myself. I don’t think coding that way would lead me to a massive performance improvement for now. Greenfield project For a second scenario, I wanted to see how it would perform on a brand-new project. I quickly set up a Python virtual environment, and asked aider to work with me at building a simple project. We would be opening a file containing Japanese text, parsing it with fugashi, and counting the words. To my surprise, this was a disaster. All I got was a bunch on hallucination riddled python that wouldn’t run under any circumstances. It may be that the lack of context actually made it harder for the model to generate code. Troubleshooting Finally, I went back to Itako, and decided to check how it would perform on common troubleshooting tasks. I introduced a few bugs to my code and gathered some error messages. I then proceeded to simply give aider the files mentioned by the error message and just use /ask to have it explain the errors to me, without requiring it to implement the code. This part did work very well. If I compare it with Googling unknown error messages, I think this can cut the time spent on the issue by half This is not just because Google is getting worse every day, but the model having access to the actual code does give it a massive advantage. I do think this setup is something I can use instead of the occasional frustration of scrolling through StackOverflow threads when something unexpected breaks. What about the Qwen CLI? With everyone jumping on the trend of CLI tools for LLMs, the Qwen team released its own Qwen Code. It can be installed using npm, and connects to a local model if configured like this: export OPENAI_API_KEY="ollama" export OPENAI_BASE_URL="http://localhost:11434/v1/" export OPENAI_MODEL="qwen3:8b" Compared to aider, it aims at being fully autonomous. For example, it will search your repository using grep. However, I didn’t manage to get it to successfully write any code. The tool seems optimized for larger, online models, with context sizes up to 1M tokens. Our local qwen3 context only has a 40k tokens context size, which can get overwhelmed very quickly when browsing entire code repositories. Even when I didn’t run out of context, the tool mysteriously failed when trying to write files. It insists it can only write to absolute paths, which the model doesn’t seem to agree with providing. I did not investigate the issue further.

yesterday 5 votes
The beauty of ideals

Ideals are supposed to be unattainable for the great many. If everyone could be the smartest, strongest, prettiest, or best, there would be no need for ideals — we'd all just be perfect. But we're not, so ideals exist to show us the peak of humanity and to point our ambition and appreciation toward it. This is what I always hated about the 90s. It was a decade that made it cool to be a loser. It was the decade of MTV's Beavis and Butt-Head. It was the age of grunge. I'm generationally obliged to like Nirvana, but what a perfectly depressive, suicidal soundtrack to loser culture. Naomi Wolf's The Beauty Myth was published in 1990. It took a critical theory-like lens on beauty ideals, and finding it all so awfully oppressive. Because, actually, seeing beautiful, slim people in advertising or media is bad. Because we don't all look like that! And who's even to say what "beauty" is, anyway? It's all just socially constructed!  The final stage of that dead-end argument appeared as an ad here in Copenhagen thirty years later during the 2020 insanity: I passed it every day biking the boys to school for weeks. Next to other slim, fit Danes also riding their bikes. None of whom resembled the grotesque display of obesity towering over them on their commute from Calvin Klein. While this campaign was laughably out of place in Copenhagen, it's possible that it brought recognition and representation in some parts of America. But a celebration of ideals it was not. That's the problem with the whole "representation" narrative. It proposes we're all better off if all we see is a mirror of ourselves, however obese, lazy, ignorant, or incompetent, because at least it won't be "unrealistic". Screw that. The last thing we need is a patronizing message that however little you try, you're perfect just the way you are. No, the beauty of ideals is that they ask more of us. Ask us to pursue knowledge, fitness, and competence by taking inspiration from the best human specimens. Thankfully, no amount of post-modern deconstruction or academic theory babble seems capable of suppressing the intrinsic human yearning for excellence forever. The ideals are finally starting to emerge again.

2 days ago 4 votes