Sasha

903 posts

Sasha banner
Sasha

Sasha

@devdotsasha

Perf Engineer. Tokyo based 🇯🇵

Tokyo شامل ہوئے Eylül 2022
351 فالونگ2.4K فالوورز
OK Then
OK Then@okaythenfuture·
People really don't seem to understand that Malaysia is basically a first world economies at this point. Salaries are even higher there for most tech roles than in London + most European cities now. Especially around Penang, which is very notable given its not the main economic engine of the country but is a huge semiconductor/engineering hub, and around Johor, which is Southeast Asia's data hub. The world has transformed massively in the past generation. The Malaysian Century.
Nikkei Asia@NikkeiAsia

Malaysia has overtaken Japan in salaries for key technology roles for the first time, driven by rising investment in the booming semiconductor industry and intensifying competition in digital sectors. s.nikkei.com/40D2o4E

English
130
946
5.3K
585.6K
Sasha ری ٹویٹ کیا
The Sales Bull 🎯 Follow if you sell B2C or B2B
Britain specialises in charging premium prices for third world results £3500 council tax just to live in your own house and the streets are still full of potholes, bins are every two weeks and you can't get a GP appointment
English
148
716
7.4K
173.5K
Sasha
Sasha@devdotsasha·
what the hell are they doing over at github
English
0
0
1
260
Sasha ری ٹویٹ کیا
Sasha
Sasha@devdotsasha·
Need to be on the road for the next few weeks / months whilst I wait for gf visa. Whats the best solution for screen real estate whilst travelling these days? Apple vision pro? Portable monitor?
English
1
0
2
653
Ebere
Ebere@eberethinkscs·
@devdotsasha Windows system -> portable monitor Apple -> iPad Pro
English
1
0
0
28
Sasha
Sasha@devdotsasha·
@ilyamiskov screen size hard to get used to, even more so in an agentic world, otherwise worth the trade offs - can handle most things you can throw at it (except cargo clippy on a large monorepo)
English
0
0
0
349
Ilya · イリア
Ilya · イリア@ilyamiskov·
As someone who's been using a 16" M1 Max Macbook Pro, I'm thinking about going with the 14" model next year. Any serious drawbacks to the 14" besides the smaller screen size? Is the Max chip too much to handle for the 14"? Curious to hear about your experience guys.
Ilya · イリア tweet media
English
211
5
369
88.1K
Sasha
Sasha@devdotsasha·
new claude use: traverse git history to find possible rationales for Chesterton fences
English
0
0
1
329
Sasha
Sasha@devdotsasha·
I’ve argued before. Once frontier models cross a certain threshold, releasing them broadly may cease to be the profit-maximising strategy. At that point, the dominant incentive is not to sell intelligence, but to deploy it to capture every industry it can reshape. I wouldn’t be surprised if we’ve already crossed the Rubicon.
English
0
0
3
328
Sasha
Sasha@devdotsasha·
I was quite being dismissive of Jevons paradox last year, but its currently giving me hands
English
0
0
1
236
Sasha
Sasha@devdotsasha·
claude is so retarded when it comes to worktrees
English
0
0
2
348
Sasha ری ٹویٹ کیا
François Chollet
François Chollet@fchollet·
Sufficiently advanced agentic coding is essentially machine learning: the engineer sets up the optimization goal as well as some constraints on the search space (the spec and its tests), then an optimization process (coding agents) iterates until the goal is reached. The result is a blackbox model (the generated codebase): an artifact that performs the task, that you deploy without ever inspecting its internal logic, just as we ignore individual weights in a neural network. This implies that all classic issues encountered in ML will soon become problems for agentic coding: overfitting to the spec, Clever Hans shortcuts that don't generalize outside the tests, data leakage, concept drift, etc. I would also ask: what will be the Keras of agentic coding? What will be the optimal set of high-level abstractions that allow humans to steer codebase 'training' with minimal cognitive overhead?
English
171
383
3.3K
320K
Sasha
Sasha@devdotsasha·
I think eventually there will be a single everything UI. Fundamentally, all a UI provides is data translation. It will be dynamically built in real-time by agents. No point building personalised custom frontends per domain, just need one. Just state the data to be presented
English
0
0
1
188
Sasha
Sasha@devdotsasha·
@RokoMijic 4GB Hetzner instance is sufficient - lags out on rust builds sometimes, but its not terrible. You can ask it to harden itself and it does a pretty decent job out the box.
English
0
0
0
63
Roko 🐉
Roko 🐉@RokoMijic·
Anyone know what a suitable machine to run an OpenClaw bot is? I don't want to run locally, I'm fine using an API.
English
11
1
8
1.7K
Sasha
Sasha@devdotsasha·
❯ cargo clean Removed 1363529 files, 706.8GiB total
English
0
0
2
306
Sasha
Sasha@devdotsasha·
Spent a ridiculous time iterating on & parallelising my workflow for agents this past week and gave myself insomnia. Least my agents are fast now.
English
0
0
0
283