• I make sites faster and cloud/LLM bills lower 🚀 • 2:43 marathoner (🔜 2:39) • Performance is my thing
Amsterdam
- Tweets 11,043
- Following 190
- Followers 72,872
- Likes 17,526
Pinned Tweet
I've found (and locally fixed) an annoying performance bug on @X web!
TLDR: go to a post with a lot of replies, and scroll up and down - notice it's laggy?
There's a component that calculates the scroll parent, and then removes the scroll listener when unmounting. But it seems we don't even have a scroll listener most of the time, so we can avoid the heavy _getScrollParent function by wrapping it in a conditional
This removes seconds of forced reflow and lag for me
Would love to get it fixed! 🙏🏻 @X @engineering @elonmusk @nikitabier
115
87
3,151
799,015
Knocked off another 28% by lowering the readiness check interval because the app now boots faster and GCP is just sitting there waiting 🙃
p99 container startup latency of a Cloud Run service down by 63% 🤠
this service scales up and down by ~3300 instances per day, so we're saving ~19 hours per day of compute time
what a nice graph
3
1
11
1,392
become obsessive about optimization 🤠
When our AI code reviewer's cost per run doubled, we dug into the token usage to find out why. Learn how we used prompt caching and intelligent model routing to slash our inference bill by 68% while speeding up reviews. cfl.re/4isj2NU
2
1
28
3,322
models are becoming so cheap
MiniMax M3.1 Flash: Space Performance + Bunny Speed
🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash.
💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.
API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens.
Model weights will arrive on Oct 23.
API: platform.minimax.io
M Plan: minimax.io/m-plan
MiniMax Code for desktop: agent.minimax.io/download
6
60
2,583
My hot take:
We should aim to streamline 90% of our life, but the remaining 10% are the most painful, eating-glass experiences that you'll remember until the day you die
I'm fascinated by the rise of frictionmaxxing as a response to a frictionless life. We engineer away inconvenience, only to discover that some of the friction we removed was part of what made life feel alive in the first place
11
3
142
8,641
Daniel Lockyer retweeted
p99 container startup latency of a Cloud Run service down by 63% 🤠
this service scales up and down by ~3300 instances per day, so we're saving ~19 hours per day of compute time
what a nice graph
9
3
69
6,233
I think the biggest flaw to the abundance mindset is the idea that you have infinite time on this planet
Time on this planet is extremely limited
Use yours wisely
5
3
129
5,652
gimme 1000tok/s models
Coming around to the idea that what I want most from the next generation of models is speed. They’re plenty good now, but I want the experience to feel more interactive. I want to say “let’s try that split vertical instead of horizontal” and for it to be shown to me <1s later.
8
30
4,106
incoming 👀
MiniMax M3.1 Flash: Space Performance + Bunny Speed
🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash.
💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.
API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens.
Model weights will arrive on Oct 23.
API: platform.minimax.io
M Plan: minimax.io/m-plan
MiniMax Code for desktop: agent.minimax.io/download
3
1
47
5,669
Everyone should learn first-aid and CPR
Thank you to the two @Cambridge_Uni medical students who saved my step-brother's life. He was running with girlfriend on Midsummer Common and his heart stopped for 14 minutes. They didn't give up on him, kept him alive til paramedics arrived. He is on the mend and should make a full recovery
3
30
5,232
Daniel Lockyer retweeted
spent the weekend making my side project's query path faster, measuring every step and re-running all of it on one machine:
443 → 2,209 queries/s through the router (5×)
2.39 → 0.31 ms of controller CPU per query (7.7× less)
no single big fix, just small changes that add up
2
3
12
1,717
Daniel Lockyer retweeted
Better pro tip: make your shit faster so none of this is needed
41
11
1,178
64,665
Another sub-3 marathon for me in Amsterdam!
Most of the race was Zone 2 heart rate, and then I pushed it a little at the end 😎
Only 3 weeks after Berlin too!
On to the next... project 2:39 resumes next year 🏃♂️
To further celebrate such an event (and to get another race in whilst I can!), I'm running the Amsterdam marathon on Sunday 🏃♂️
No time goal... having fun is the goal!
It's my home event and I know a lot of people doing the race, so it's great to be to do another marathon 🤠
59
3
409
27,686