I am able to get pretty good writing with the right chain.
<Provide my writing>
=> highlight unique key phrases that are unique and surprisingly style that wouldn't be written by an llm
<provide a LLM re-written version on the same topic>
=> Write a "voice guide" that highlights why my voice is unique and how to produce similiar writing
=> Write a "How to identify slop" that does the same
=> Then use it with the given guides to write new ideas
```
where:
=> is a prompt, and <> is context
@viccpoes I think it would perform even better in a different poll.
Head to head is not real world use.
Across many iterations and prompt refinement it crushes nano banana, gpt image, etc..
@csmisko is this streaming from a server?
And im suprised it doesn't fly apart with enough heat, how much heat does it take to break the molecule into atom?
NERF's are one of the ground shaking papers in the field.
Most ML papers are boring architecture tweaks, very few change how we think about new conceptual spaces.
Countable on one hand.
Congrats @BenMildenhall and @_pratul_! ACM gives Grace Hopper Awards to under-35-year-olds for *one specific thing*, here it's NeRF.
Past winners include Donald Knuth (for "The Art of Computer Programming"), Steve Wozniak (for Apple), and @martin_casado (for SDN). Good company!
@aleabitoreddit@StormDirac So what evidence did you have to see this coming? You never really explained why you had so much conviction that they would be in deals? What docs did you read.
@VictorTaelin that's awesome. Is there anything you can learn from the compiler? Like modern JS gets all it's speed from JIT compiling.
Maybe you can hot spot precompile the bottlenecks in the interpreter.
@citrini Someone to go track down all of the hyper scalers GPU orders and trace where the actual chips are , how many are sitting in warehouses, how many are in racks unpowered, how many have power and are serving. Accounting for every purchased card.
@levelsio Luxury money used to mean they bought the best designers and sourced the best manufacturers. But now with globalism really ramped up and our supply chains being so mature. The companies with scale drive cost down and quality up and luxury can't compete with scale.
New in Claude Code (research preview): dynamic workflows.
Claude writes an orchestration script on the fly, then spins up a large fleet of coordinated subagents in parallel to take on your most complex tasks.
Use the word "workflow" in a prompt to get started.
@sidbid This is so cool and it's very good at self utilizing this, I requested branched tree rollouts of cascading prompts, and it wrote a workflow and used it and I didn't even know it was out yet.
Hmm... Interesting. I'm not so sure it's the same here, because you can still write and deploy code and ignore the code itself.
Chess is fundamentally simple where if the robot is good at chess you can study it's moves and the moves are the game itself.
But in software there is the code of the software, and the behavior of the software.
And software engineers are paid for the behavior not the implimentation.
I think software engineers CAN become much better if they focus no the craft and read and understand what the models are writing, but I think we'll see a real bifurcation were most people become less experts and more just supervising.