pianist, organist, harpsichordist, computer programmer
Joined January 2025
- Tweets2.1K
- Following139
- Followers189
- Likes9.1K
Pinned Tweet
Just got the piano tuned - the first part of Couperin's La Couperin to celebrate :)
I have proper microphones now but still need some cables and to get everything set up. I'll start recording full pieces soon!
barafostus dreame retweeted
but data-oriented design is not mere code optimization, it's a design principle, and with modern hardware and compilers, almost all *code optimizations* are design decisions that live at the architectural layer. i would understand if you were referring to instruction-level micro-optimizations, which we don't do anymore because compilers are good(sometimes). all memory layout optimizations are architectural and should be thought of from the start, pre-profiling. if you already understand the general latency of cache misses and the machine's execution behaviour , which is deterministic. you already know how to design they architecture around that, in a coherent way. good hw/sw co-design
barafostus dreame retweeted
I disagree. Performance engineering is designing from day one around real-time performance guarantees such that you don’t ever have to do a deep dive in performance profiling later on to try to identify and hammer out each issue one at a time. Such an approach hits diminishing returns fast, dramatically increasing the amount of effort for ever small gains each time.
You ultimately run into an issue where your performance issues are death by a thousand little cuts and you need a full rewrite with an entirely new architecture and interface to achieve meaningful improvements. However, by then it’s too late as you’re locked into the existing interface that everyone has already built content upon and it’s your primary revenue stream.
All of which goes to say that being proactive about performance from day one is way more important than trying to be reactive to it after the fact, as you’re suggesting.
Also flame graphs only show detailed performance issues if you’ve architectured your codebase to be a bunch of many small functions, which is actually horrendously annoying and slow when trying to optimize an existing codebase. It also poorly handles capturing occasionally spikey performance.
i'm seeing so much performance slop in the timeline.
> "How can we use data-oriented design principles"
> "Can we turn this pointer into an index"
Perf eng is get CPU flamegraph, maybe some heap , disk , and network profiles.
Make biggest bar on these profiles smaller, if not move to the next biggest bar.
barafostus dreame retweeted
Because a binary tree is the most fundamental data structure. Any programmer who is scared of binary tree operations does not know any data structures at all, and thus doesn’t know about algorithms either (since data structures are what give algorithms their speed). Thus he knows approximately 0 computer science and the interview can safely end.
barafostus dreame retweeted
Chesterton presents an alternative explanation:
"What he gives up must be some simple and universal thing. He will give up beef or beer or sleep–because these pleasures remind him that he is only a man."
An opinion I hold about command-line and other interfaces that might be uncommon is that optional parameters are bad and should be avoided as much as possible, ideally entirely. I think it's a good discipline to be forced to think about exposing the least program surface area. Now you have to think about design.
barafostus dreame retweeted
The modern design principles would not be able to design a scientific calculator UI. It would be like a dumbass 6 button gadget with a touch screen wizard. There would be debates on Twitter whether numerical input should be up/down or left/right scrollers. Most of the criticism would be about some tiny 1px misalignment and how that's bad design. There would be concept showcases and improvements like adding tweening animations, one comment would say "Oh I love that bounce! ✨". The general concept of a shift button would be entirely alien, let alone two of them. LSK or Line Select Keys (the blank top row buttons) would be considered unintuitive because it has no labels. The new Calc Pro Ultra Max would tout itself as the professional choice, with a grand total of 8 buttons.
ALT Source: https://www.reddit.com/r/calculators/comments/1qzmr5r/long_lost_hp_48g_siblings/
barafostus dreame retweeted
Reddit call was absolutely fascinating. Everyone should listen to it as a bellwether for where the internet is going
- AI Mode is increasingly default Google experience
- AI Mode provides rich info and data in line, less incentive to leave search
- AI Mose incentivizes dialog, Google wants to keep you engaged to serve more ads at each turn
- Google will milk search for as much revenue as possible as they transition from a search company to a cloud service and chip company
- As traffic from humans becomes more scarce this will put Google in an increasingly adversarial position to the rest of the internet
- This is a massive headwind to all media type companies who are not destinations
- This is a massive headwind to all content-driven go-to-market motions especially in the SMB. You have seen companies like HubSpot and Monday.com get crushed by this funnel collapse
We are entering the end game - a knife fight for the remaining and dwindling volume of human traffic on the internet
I wonder what the next step after all this is.
Robot cleans up swathes of open human mathematics. Maybe not everything straight away, but a lot.
Is robot maths coming next? The bots think about maths a lot and tell us about it and we try to understand what the hell they're up to?
WHAT THE FUCK
Here's a direct link to OpenAI's PDF claiming a refutation of ETH. In particular, it's claimed that 3SAT on n variables is in about exp(n/(log n)^{1/2+o(1)}) time
github.com/openai/math/blob/…
The Bun people being LLM maximalists is hilarious to me given their entire product is a JavaScript runtime. This is like a horse-and-buggy salesman extolling the power of the combustion engine.
With current LLM capabilities, JavaScript no longer has any reason to exist.
