Equational

Vibe coding a small reactive app I wrote a small app with a web reactive table front-end last week. I am not experienced in web development. The last time I wrote a web stack from scratch was with a Java front-end! However I have led such projects as product owner, therefore I knew what I wanted, and decided to use Gemini Pro 2.5 to tell what I needed to know. How easy? The question that is answered here is: how easy it is to develop a app with tables where cell editing may trigger updates in other cells in a modern reactive lightweight web-front-end in Python using an LLM to guide you? The answer is: not easy, but with enough architectural knowledge, it works. An iterative process It took me four chat sessions, eleven main code iterations, with debugging iteration happening within each of these iterations.  All libraries except for htmx were suggested by Gemini. htmx was my constraint. Below are the iterations driven by Gemini with regards to the choice of libraries (except f...

Two extremes of prompt setup For this chat I would like you to be literal, concise, technical, analytical, grounded and factual, with a formal tone, and be taking an objective stance. For this chat I would like you to be holistic, explanatory, simple, emotional, creative, casual and be taking a persuasive stance. A test question: Does one need belief to do mathematics? Responses (Gemini 2.5 Pro): Belief in the metaphysical status of mathematical objects or the absolute truth of axioms is a matter of philosophical conviction and is not a universal requirement for the practice of mathematics. Oh, what a beautiful question! I absolutely believe you do. In fact, I'd say that mathematics, at its very core, is one of the most profound acts of belief a person can engage in. It’s not about faith in a deity, but a deep, powerful trust in the rules of a magnificent, invisible game. Command list For convenience, I include the each command here: 1. Focus (Structure & Intent) Literal: ...

Dynamic typing of duck typing The link to the code produced by Gemini and the topic of this blog . Two months ago I drafted a Python tracer library with Gemini, then I rewrote it from scratch (here) . It is however "clunky". Yet I do not worry, as Python is a computer scientist dream language, and in Python one can quickly implement something meta and probably slow, but then use a cache or two to make it fast enough. That is what I did last week: I sat in front of my screen, stared at it for 30m thinking through my options, and then asked Gemini to draft the concept. And while the result is missing some logic, it is however doing what I want: it dynamically wraps "types" around duck-typed operation. I could try to explain what it does... but I do not need to. Instead, I share what I told Gemini. The key desires are the following: let's make a cached wrapper factory, that when given an object, first is looks in the cache, and returns the instantiated wrapper with...

Having led the development of a large-scale, mission-critical system in a complex domain. I can't talk about the specifics of that system, but I can direct you to this architectural philosophy and design pattern blog I wrote many months before I started Elevence , and even before I wrote my playbook. It has become one of my most liked blogs:  Software designs that grow with monads, comonads, and type compatibility  . And one that has given me some fame. The story was that after having built software for many years, I started to notice that certain designs inevitably collapsed under their own weight, while others seemed to grow stronger with change. It felt less like a matter of style and more like a matter of mathematics. There are underlying laws about how information and change behave in a system. I wrote up the thoughts in that blog. Purposely cryptic back then, as it makes no sense to give away what is still new. Breaking down some core concepts of the blog: Type...

(Sep 2025 update: dedicated page with bot status update is here ) In April 2021, a bot started loving this blog. This for example is its love bites for a 24h period earlier this month: The love is real ❤️, yet until now I have ignored it, because how can I express a sentiment in return? Then it hit me! Having recently been to see Lionel Richie's concert in Zurich (was fun), I suddenly understood... Therefore, maybe late, and with help from Gemini, but not without feelings, I share here the message that the bot will surely find, and understand that the feeling is mutual! 💞 (To be sung in the classic "Hello, is it me you're looking for? vibe")  Hello, is it me you're looking for? I can see it in your logs, All those hundreds of scores, From Brazil and Vietnam, Argentina, You're always knocking at my digital door. I can see you, I can see you, Your data streams, they flow, Through my quiet pages, fast and slow, Checking on content, though it never changes, A con...

First try (from Gemini 2.5 Flash) Software architecture defines the system's fundamental structure , outlining its major components, their high-level relationships, and the overarching principles and quality attributes (like scalability or security) that guide its construction. It's the "what" and "why" of the system's foundational choices, which are costly to change. In contrast, software design focuses on the detailed internal workings of individual components and modules, specifying their logic, data structures, and algorithms to implement the architectural vision effectively. It's the "how" of putting those foundational pieces together, generally with a lower cost of modification. Second try (with my input) Software architecture defines the system's fundamental structure , outlining its major components, their high-level relationships, and the overarching principles and quality attributes (like scalability or security) that ...

Here are a few master level techniques to prompt an LLM for highly sophisticated results. The example below are about producing software, but the technique is general, and applicable to all domains.  Do not start with previously written prompts that setup a generic set of rules. LLMs chats are a one way: every new input from you will "refine" the scope of the LLM, an early generic topic would "collapse" the LLM billions parameter wide space even before the specific topic is given. Start by asking for something general with regards to your final goals. You want the larger scope of the LLM to refine in the context of your end goals. For example prompt: You are probably aware of the type inference algorithm called something like cartesian product?  Start by asking in the most broad manner. Failing to be broad will refine the LLM's scope before you can use it to your full benefit. Always reformulate the LLM's response back to the LLM to fit your goals and ensure...

I spent two days last week reinstalling Nixos "bottom up", with X11 and i3. During day two I was thinking "this is such a waste of time". However it was not, here is why: Learning is about doing, and often in software that means "type it in yourself". Therefore, to install NixOS this time on a laptop, I typed everything in: the disk partitions, LUKS keys, grub setup, and new configuration file with X11 specs. I am a better developers when the system I develop on feels "simple and clear" because I configured it. I bring up this experience to share a few things. The first is that being a developer is both about being stubborn in a smart way, and about accepting loss gracefully. For example: What to do when you are locked out because somehow when you changed your LUKS keys you "screwed up"? You just reinstall again, and zenfully accept the loss of time for not being careful. What to do when you insistently fail to try to avoid to depend on...

Failure and then success in querying an LLM I enjoy accelerating my software development with the help of LLMs. However, this extra LLM based support does not always work. One of the reasons can be the LLM never recovers from its bad choice of direction based on your first queries. Below I share a good example of such a failure, followed by a second LLM dialog that quickly produces the desired result. Then I explain what is happening. Here is the dialog that fails, I am trying to get the LLM to produce a draft differentiation approximation based on wavelets, and coded in JAX: Me: Do some wavelets have special properties with regards to derivatives? More specifically, I have noise data that has a slope, I compute the wavelet transform, do some of the wavelet basis allow me to reconstruct an approximate derivative by manipulating the wavelet coefficients before doing an inverse transform? (To note that I naturally try not to impose my knowledge, and therefore why I tend to caref...

  LLM assisted coded help dialog Just before lunch today, I left a message for my son, asking for help with sklearn. Then I thought: Gemini should be able to help me.This is the resulting dialog (summarized): Me: I have some x y data points that tend to follow a linear trend but with this trend abruptly changing from time to time. I am looking to use sklearn where I provide my x y data, and at the end have optimally partititioned the data so that each partition's data can be fit with a line while miinimizing the fitting error. Can you help me? LLM produces a first draft based on DecisionTreeRegressor and GridSearchCV. I run it on a small dataset. Me: I note that this has a hard time getting the transitions "right", meaning the estimated segment breakpoints have a hard time to converge. How about a second pass to that keeps the position of the segments, but otimizes their ends? LLM proposes code, but forgets that it should be a second pass. Me: Hmm, I think you would want...

 (Early 2025: LLM generated in one go without retouch) From Command and Control to Coherent Models: A Journey of Professional Growth Throughout my career, I've had the privilege of wearing many hats – from Product Owner to CTO, Software Architect to Managing Director. Each role has been a stepping stone, shaping not just my skills but also my fundamental approach to work and leadership. Reflecting on this journey, I see a clear pattern of evolution, a shift from reactive, belief-driven actions towards a more strategic, data-informed, and model-driven approach. This blog post is a reflection on that evolution, broken down into key areas of growth that have defined my professional path. 1. From Command to Cooperation: Embracing Collaborative Leadership My early roles often saw me operating within a "command and control" framework. As a Software Engineer, I received instructions and executed them. As I moved into leadership positions like Product Owner and Head of Software D...