Discovering new math and inventing new money Science, economics, programming, and general weirdness. dev @Indexedex Original Olympus developer.
Sixth Dimension
Joined June 2019
- Tweets18.5K
- Following6.7K
- Followers7.6K
- Likes31.4K
Pinned Tweet
The downtofinance X account was hacked.
The site is still safe. I still control that.
I had hired someone I know to handle the X account. They fell for a phishing attack. I know them, they were just being dumb.
I was planning to rebrand to Indexedex when I deployed to other chains. So I'll just do it now.
Site update and migrating the staking to the fee accrual DETF is coming today.
Official X account for this rebrand is @Indexedex new site is indexedex.com
Old site is still up. I will be changing that page to a notification and link to the new site.
Maths are WMDs.
The left wants 20 years in prison for software engineers building AI.
That is more than kidnapping.
More than twice child [redacted].
Four times drug trafficking.
The same band as murder.
Sanders and Casar compared linear algebra to rogue nuclear weapons.
Proposed max. Not law. The chart is first-time federal sentences.
Working with AI is a lot like being a Team Lead or CTO.
Each model has it's strengths and weaknesses. Treat them like different types of staff.
@ollama Is your junior dev squad. A mix of talents. The harness and processing resources defines their behavior more than the specific model.
@AnthropicAI Opus and Fable are your standard dev. They'll get it done, but will make some weird decisions along the way. And they'll cut corners. They value delivery over correctness.
@grok is your research and math minion. The supreme autist. And like any autist, it's a bit naive. Tends to believe what it reads over verify with code.
@ChatGPT Astra is your senior dev. Writes tight code, obeys the style guide, follows the project patterns. And it will grind on solving the problem and proving it works with tests. But it doesn't want to make any design decisions if it can avoid it. Wants to delegate all decisions to the human.
The quality of product is mostly defined by your prompt. If you write a good PRD, and detailed implementation plan any of them will complete the task.
Use the frontier models to do the research and write the plan. Then cheaper agents can deliver at comparable quality as a fraction of the cost.
Lord of Light by Roger Zelazny is the best book ever written. Fight me.
en.wikipedia.org/wiki/Lord_o…
We need better language to describe emergent effects.
The only thing more terrifying than a cabal secretly controlling the world is that they're isn't.
This is why I invented the Reverse Harbinger Auction.
None of the AI bulls can give me a good response to this question:
Why does CheaperInference, a service that buys excess compute at a discount and then sells it to end-users at a rate below standard API rates, exist if we don't have enough compute, resulting in us needing to build out datacenters so much?
Someone make it make sense.
It just seems like the most obvious "yeah, it's a bubble" moment for me.
It's easier to train a smart dog.
