Technologists love to be sure. "That can't happen.” “Oh, it's inevitable.” “That's impossible.” “It’s a power law." But that certainty is a shield against the uncomfortable truth that reality is fundamentally unpredictable and confusing. To some, it’s a point of pride to be right, and to be convinced of it. But there’s nothing wrong with being wrong, sometimes. The most dangerous systems aren't the ones that can fail, but those that convince you they can't.
When everyone repeats the same story with unwavering confidence, look for what they're collectively missing. Nothing is a panacea; tradeoffs are built into reality—despite what X posters and Hacker News commenters suggest.
Code speaks with confident certainty and those who write it for a living often do the same. Yet errors and type signatures lie, null pointers exist, 500s are mixed up with 404s. Our industry endlessly debates the one true way to solve problems and the software we produce declares its ignorant certainty on every interaction. Tacitly, we know that the code doesn’t tell the whole story, which is why debugging is so taxing. Sometimes I wish the code would say ‘I don’t know’ instead.
Code, Command lines, REST APIs, and LLMs force multidimensional thoughts through an expensive translation into certain terms1. As the operator, you have to dedicate most of your capacity to that articulation, rather than direct ideation. When you translate a thought into an expression, you're using someone else's APIs, ideas or phrases to do so.
Digital direct manipulation interfaces emerged2 precisely because humans don't intuitively think in terms of precise commands, linear steps or perfect articulations. By the time you can say what you mean, most of the thinking has already happened. Why else would it be so difficult to put our most meaningful thoughts into words?
Thought is the process of grappling to comprehend the unknown in terms of the known, an endless spiral. To master any pursuit you must abandon certainty. A master is never certain—they're constantly revising, forgetting, and rediscovering3.
Reclaiming Agency
Tech culture celebrates “building”, but real innovation begins with demolition—tearing down what you were certain about yesterday. I have no interest in comfortable narratives about technology or computation being inherently good. I would rather use a given technology for unintended purposes.
If we’re going to let LLMs dominate the future of code then we owe it to ourselves to ignore the tired narrative of what this technology is, how it must be created and delivered to people.
If we’re going to spend all this time talking to our computers they should not infantilize us or constrain our thinking to the known. I grew up teaching myself whatever I wanted using the internet and formed my own opinions of the world. I’ve always sought after contrasting perspectives and impressions because it’s in the juxtaposition that learning occurs.
The main value I see in LLMs is for curious minds to more rapidly explore a greater surface area of perspective. Except, this cuts both ways. Information addiction is real and while an ‘aha!’ moment feels good... it often doesn’t lead anywhere useful4. How can technology help us build tacit knowledge? Such technology must help us wrangle the unknown, collaboratively.
I am still processing if, how, why and when to use LLMs to the greatest effect. What I’ve learned so far is that I disagree with pretty much everybody… As usual. More soon.
✌️ Ben
Stuff I’ve Been Thinking About
Reality Check - Ed Zitron
The Wave in The Mind - Ursula K. Le Guin
The case against conversational interfaces - Julian Lehr
Google shattered human connection - André Staltz
System 3 thinking - Educating Silicon
Yes, I am saying words are certain. You may not be certain about them, but they are symbols that others can interpret with certainty
Obviously physical tools are the origin of direct manipulation, further show that language is auxiliary
How does that compare to your computer?
A frustrating but familiar aspect of creative work. Sleep on it, the idea won’t seem so amazing tomorrow.
(yeah, you read it right)