@ForrestPKnight Well, here is another data point: I work at HashiCorp (an IBM company now), and have not hand written code for months. I still "hand"-review though.
@AnthropicAI I see that Opus responds to queries with JSX, and https://t.co/Jnbw0ijyk6 visualization. What can we do to make recharts better for agents?
With LLMs I am able to generate, validate and share ideas quicker than the organization can digest. I need to stop myself and focus. There is no value in producing quicker than can be digested. The bottleneck did shift - but it did not disappear.
@tottinge I do not see that at HashiCorp. To me, the main thing is that I can now explore ideas that I would previously not find time for.
On the negative side, I see that the default answer now is to rewrite from scratch rather than reuse what exists - just because one can.
@nexxeln@opencode Maybe I am missing something - but was that not already possible? I would tag subagents to review a PR, and then have multiple parallel agents reviewing.
@DominikWKH Selbstwirksamkeit bezieht auf die Überzeugung und ist daher beinahe schon ein Gegenteil zur Agency, bei welcher es um die Handlung geht.
Ich befürchte wir haben hier einen jener Fälle, bei dem kulturelle Eigenheiten eben nicht übersetzbar sind.
I think we are witnessing the biggest explosion in software creation in history.
New website creation is up 40% year on year. New iOS apps are up nearly 50%. GitHub code pushes in the US jumped 35% and in the UK around 30%.
All of these metrics were flat for years before late 2024. The entire graph looks like a hockey stick.
You no longer need a six month runway and a dev team to ship something real.
We see this in our metrics as well!
People who never wrote a line of code are building and launching apps.
The barrier to building software just disappeared.
What matters now is knowing what to build and the taste to build it right.
Building the wrong thing is now easier than ever. I see many people building complex systems because they can. First principles matter: Instead focus on the value proposition and validate.
@roy We can not expect consistency from the stochastic parrot. Adding spaces to a prompt changes the response. It is an amazing tool, but we are misled to anthropomorphize and expect from the parrot what we expect from other systems (humans) that generate such word clouds.
@NeoliberaleA Kosten > Nutzen. Mit steigendem Wohlstand sind die Opportunitätskosten deutlich gestiegen. Wer heute Kinder hat verzichtet auf mehr als je zuvor. Der Staat erhöht weiter die Kosten durch Mietpreisbremse (Verknappung von Wohnraum) und Gurtpflicht (>3 Kinder braucht ein neues Auto)
Inspired by this post, I looked into this a tad further. Minor changes in the prompt result in different answers, there really is no consistent spirit behind the curtain. https://t.co/qWsTAtrm6R
I had ChatGPT and Claude discuss the highest value books until they both agreed to 3
They decided on:
An Enquiry Concerning Human Understanding — David Hume
The Strategy of Conflict — Thomas Schelling
Reasons and Persons — Derek Parfit
@BecomingCritter I am surprised to not see a book on evolution, and as I ask the same question (Opus 4.5), I get:
- The Republic by Plato
- The Origin of Species by Darwin
- The Wealth of Nations by Adam Smith
Which makes me wonder: Might our personal chat history be reflected here?