Solvitur Ambulando - 'It is solved by walking' - is a Latin maxim. It has a modern counterpart, Solvitur Colloquendo - 'It is solved by conversing'. The parallel is deeper than mere wordplay.

If you code with AI, you have already experienced this but may not have named it. Naming it changes how we approach it.

Solvitur Ambulando, The Ancient Method

The phrase is attributed to Diogenes, but my introduction to it was in one of the letters to a disciple by Sri Aurobindo. The solution to a certain spiritual impasse was to take the next immediate step, however small.

This act of walking to think or solve something is quite common amongst those who ponder mental abstractions, mathematicians, poets, philosophers and even programmers.

In Yogic practice, walking is a means of externalising the inner meditative state, to infuse some objective dynamism to the subjective equipoise of meditation.

So keep this 'solving' in mind, and you can see how our new maxim matters.

Solvitur Colloquendo, The New Method

Programming with AI is conversational. We do not write a set of watertight instructions in English to a compiler. Our conversation is with AI. We describe our goal, our intent in attempting it and our constraints. AI gives options on how we may approach it.

Take for example one of my projects after reading the Tiny Stories paper on SLMs. So I wondered, can a 20 million parameter model learn complex topics when the amount of training data is very little? I wanted to take Sri Aurobindo's Complete Works as the dataset for training this model.

So I began chatting with Claude Code (CC). Gave it details of my aims, the data extraction challenges I faced earlier, chronology facets in the corpus, the small dataset size and how I was not happy with the evals I had in place.

CC started addressing each issue I had; step by step we discussed every step of the entire training/eval pipeline. CC was a patient and knowledgeable co-thinker, explaining what I did not get, suggesting options where there were limitations. We mapped out my unknown unknowns, this was the key for me.

I began with a curiosity to create a single small language model, but the constraints of this problem led CC and me to plan for a suite of models, with varying configurations. I have learnt a bunch of new concepts and ideas, am yet to be done with this exercise but without CC the project would not have the significance it has now.

The Deeper Parallel — Why Both Work

The deeper parallel is this, as walking is to the body, talking is to the mind.

Mental knots, wrongly or poorly wired ideas need intelligence to untangle. To find if there is a gold nugget in there.

Our ideas are often too raw, as mine was.

But in chatting with CC I started seeing the contours of the idea better and could get to a meaningful and pragmatic problem formulation. An objective intelligence had helped prune the rough edges and helped discover blind spots.

AI can give you all the edge you need, if you converse with it in an open and frank curiosity.

So, what changes if we name this phenomenon?

What Changes When You Name It

Once we recognise AI assisted development as colloquendo rather than "using a tool", our perception of it, our relationship with it changes. There is a definite shift in our cognitive poise.

Focussing on prompts makes our cognitive machinery misperceive the nature of AI. This is no slot machine where you slide in a prompt-coin and out comes code, presentation or a report.

With colloquendo we acknowledge that AI has more to offer than what is harnessed thus far. That opens us to more possibilities of engaging, exploring areas yet new to us. You could still command it for favours, but you miss the point and are blind to its potentialities.

Naming the phenomenon could help those of us who are yet to experience this new mode of problem solving, and to those who know it already, it can help optimise their cognitive process to utilise AI better.

If there is one take away from this post, let it be this. Solvitur Ambulando for issues in Yoga, Solvitur Colloquendo for problems of the craft.

It is a rich day, when we get to do both.