L I V E W I R E retweeted
How we made our frontend CI ~4x faster, from 4 min to about 1 minute, while I was at the gym.
Saw this tweet, opened Claude on my phone between sets, and had it ship four changes:
- Cache node_modules itself, not just the pnpm store. The store lives on a different disk, so every run was copying every file anyway. Install 52s to 8-16s.
- Typecheck with TypeScript 7. It's the compiler rewritten in Go, so it runs native and in parallel instead of single-threaded JS. 22s to 6s.
- Shard our Vitest unit test suite 4 ways. Most of the time was jsdom setup and imports, not the tests themselves, so it splits almost perfectly. 2m31s to 33-54s per shard.
- On PRs, Vitest only runs the tests the diff can reach. Touch config or the lockfile and everything runs. Typical PR tests now take 7-12s.
It's pretty optimized now. Most of the remaining minute is runner setup and install, and squeezing that further means something like moving to Bun, which is a much bigger lift. This was the cheap stuff.
CI runs on every PR and every push, for every dev, every day. When it's slow it's a constant drag on the team, and when it's fast the time saved compounds. Less than an hour of my time, from my phone, is a pretty good trade.
Our full CI pipeline is now sub 1 min. Here is what I did:
- Vitest -> bun test
- tsc -> bun check
- oxlint -> bun lint
- oxfmt -> bun fmt
- shards filled by recorded test time
- PRs run only the tests the diff can reach
- backend builds the app once per process
- lint, typecheck, build and tests all in parallel
200+ hours of waiting saved every week
L I V E W I R E retweeted
Updated: prompt to get your CI under 1 minute. ⚡
Our full CI pipeline is now sub 1 min. Here is what I did:
- Vitest -> bun test
- tsc -> bun check
- oxlint -> bun lint
- oxfmt -> bun fmt
- shards filled by recorded test time
- PRs run only the tests the diff can reach
- backend builds the app once per process
- lint, typecheck, build and tests all in parallel
200+ hours of waiting saved every week
L I V E W I R E retweeted
words are failing you during ai interviews or when taking assessments abi 😂
well, one of the fastest path to getting instant offer isn’t endless applications… it’s performing well in assessment.
you can usekestrel.net ai to ace your interviews and assessments anytime
L I V E W I R E retweeted
the biggest bottleneck in coding with AI is the Github CI.
builds and tests take anywhere from 30-60 minutes every single time, and you end up wasting hundreds of hours every month.
what's the best alternative for this??
L I V E W I R E retweeted
if you truly want to be hired on mercor
let me give you some advice.
take this weekend and complete these two courses on mercor
once done, pick up your cv and update your professional certificate
from monday, stop applying to skill based roles, instead focus on generalist roles
there about 5 generalist roles opened and i shared them in my previous post, would drop link in comments
with these courses done, your skills and certificate updated on your cv, you stand a better chance of getting hired…
L I V E W I R E retweeted
Valuable Tip:
Tell Claude not to use "Computer Use", let it edit the video in the background. let capcut be closed whilst prompting claude. And give claude the access to the download folder or folders where the assets are in
Claude will do everything from scratch in the background, without you having to wait till it's done... and you will be able to keep doing other things with your PC.
That is way better than this.
You're welcome.
Claude AI, edited this complete 31 minutes video perfectly, I didn't generate a single, image nor video.
I didn't manually generate voice over, sound effects or background music.
I must be very honest that Claude did the editing better than I could have done it, even if i spend 6 hours editing myself.😂
All I was simply doing is directing it on areas to fix, when I discover a segment that needed fixing.
Editing is one major challenge most face YouTube creator creators face, even myself, now i end up creating videos in peace, while use my time to do other things.
This is why we call it YouTube automation. I love this soft life!
L I V E W I R E retweeted
We shrank 1 GB of Wikipedia to 105.39 MB, losslessly.
Introducing tufazip: the best entry on the Large Text Compression Benchmark using a neural network as its sole predictor. Everything above it mixes a transformer with hand-built models.
Full post 👇
tufalabs.ai/research/tufazip
L I V E W I R E retweeted
💡 12 Claude Code skills I use to ship iOS apps. 6 of them are made by Apple
give this to your agent, tell it to improve the skill and run
sauce-ops.appskins-studio.wo…
L I V E W I R E retweeted
How to Apply for UN Volunteer Roles
I promised to show you step-by-step how to apply for UN volunteer roles. I will not only show you how to apply but also, show you an easier route to better position yourself for these roles.
Thread 🪡 1/6
L I V E W I R E retweeted
being competent at your job means more than just showing up, completing assigned tasks, and collecting a salary.
here are 5 effective ways to become competent and know your onions in the workplace:
L I V E W I R E retweeted
Sometimes the hardest part of Mercor is getting your first opportunity.
This person was scared about their situation, struggled with technical interviews, and English wasn't even their first language.
Instead of giving up, they started taking the interviews and assessments seriously and used mock interviews to prepare.
Then, within one week, they received 3 offers.
If you're applying and nothing is happening yet, don't assume it means you're not good enough.
Prepare better. Take the assessments seriously. Keep applying.
Your first offer may take longer than you expect, but it can happen.
@mercor
L I V E W I R E retweeted
/make-logo skill is live on design.how
tell it about your company and it’ll find bunch of references from are.na and build you a brandbook
pls send me what you make so I can keep improving the skill!
L I V E W I R E retweeted
I just refreshed my French proficiency exam study doc! If you’re looking for a free, high level resource to help you prep for the TCF exam (for Canadian PR or otherwise) you can start here docs.google.com/document/d/1…
L I V E W I R E retweeted
Just shipped otpkit — an open-source MCP server for temp email & SMS.
Create disposable emails, get free SMS numbers, auto-extract OTPs & verification links. Zero API keys.
github.com/SecFathy/otpkit
found some Bug Bounties!
1. usual labs- 16,000,000 USDC
Bounty range: $50,000 - $1 million
audits.sherlock.xyz/bug-boun…
2. Arbitrum- $2,000,000 pool
Bounty range: $1,000 - $500k
immunefi.com/bug-bounty/arbi…
3. Kamino- 1,000,000 USDC
Reward range: $50,000 - $100,000
audits.sherlock.xyz/bug-boun…
4. Redstone- $250,000 pool
Bounty range: $50 - $250,000
hackenproof.com/programs/red…
5. Ticko- $200,000 pool
Bounty range: $50 - $200,000
docs.ticko.xyz/resources/tic…
6. Rhea finance- $250,000 pool
Bounty range: $100 - $250,000
hackenproof.com/programs/rhe…
7. Gold fish- $250,000 USDC
hunt.certik.com/explore/gold…
8. Aave V4- 2,500,000 pool
audits.sherlock.xyz/bug-boun…
Bug bounties are something that never ends in this space🫠🫠
























































