flipflipyong Profile Banner
Philip Yong Profile
Philip Yong

@flipflipyong

Followers
60
Following
621
Media
20
Statuses
269

building agents for the vibes šŸ¤™šŸ»

Kuala Lumpur
Joined May 2012
Don't wanna be here? Send us removal request.
@flipflipyong
Philip Yong
12 days
Is @Cloudflare down again lol
0
0
0
@flipflipyong
Philip Yong
14 days
Recent acquisitions go brrrrr
@bunjavascript
Bun
14 days
Bun is joining Anthropic! https://t.co/06NlXDZhNE
0
0
0
@flipflipyong
Philip Yong
20 days
I’m thinking of getting a 3D printer to make random things for fun, what’s good 3D printer? I feel like I’ve heard of Bambu Lab A1
0
0
0
@flipflipyong
Philip Yong
25 days
No more slop.
0
0
0
@flipflipyong
Philip Yong
1 month
The developer life is figuring out how to build a solution and build it. Then ends up learning a ā€œbetterā€ way the next day and then wanting to rewrite the solution. And if the dev does rewrite it, the dev might find another ā€œbetterā€ solution a few days later… That’s me
0
0
0
@flipflipyong
Philip Yong
2 months
I love building with @payloadcms Makes CMS feel fun again
2
2
18
@flipflipyong
Philip Yong
3 months
Claude's back
0
0
0
@flipflipyong
Philip Yong
3 months
Claude Code’s down 🄲
1
0
3
@Pokemon
Pokémon @ Lumiose City 🄐
5 months
We are deeply saddened by the passing of James Carter Cathcart, the voice behind Professor Oak, Gary Oak, James and Meowth of Team Rocket, and other iconic characters for the PokƩmon animated series over many years. In addition to his voice work, James provided over 700 episodes
746
13K
83K
@flipflipyong
Philip Yong
5 months
Why are we quoting ChatGPT now lol
0
0
0
@flipflipyong
Philip Yong
5 months
It's still pretty good tho!
0
0
0
@flipflipyong
Philip Yong
5 months
The problem now is that I didn't have sunlight in my data set so the model is struggling when he goes into the sunlight. Gotta annotate more and retrain the model!
1
0
0
@flipflipyong
Philip Yong
5 months
And it got pretty accurate results!
1
0
0
@flipflipyong
Philip Yong
5 months
Had a model trained on the annotated data
1
0
0
@flipflipyong
Philip Yong
5 months
I 'm recording a live stream of my puppy at the pet hotel but a whole day's worth of videos is too much to watch. So I decided to train a model to identify my puppy and help me to identify which timeframes he shows up on. I annotated 20 mins worth of footage
1
0
0
@flipflipyong
Philip Yong
5 months
Next step to optimize valuable watch time would be to train a model to recognize my puppy and identify which timeframes he appears so I don’t have to shift through a whole day’s worth of videos. After that, highlight if he’s doing anything wack
0
0
0
@flipflipyong
Philip Yong
5 months
Best part of being an engineer is being able to bend the rules and do things people usually don’t think are possible. I have my puppy at a pet hotel and they have live streams but no recorded videos so I wrote a script to watch the stream, record it and upload it to my GDrive so
1
0
1
@flipflipyong
Philip Yong
5 months
It’s so fun that with tools like Bolt and https://t.co/sNSGeui9FJ, anyone can create that one small/fun idea that you just desire to have. My wife has been manually recording our puppy’s eating, pooping and peeing timers on WhatsApp. She includes images too and now I just vibe
0
0
0
@karpathy
Andrej Karpathy
6 months
+1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window
@tobi
tobi lutke
6 months
I really like the term ā€œcontext engineeringā€ over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.
529
2K
14K
@flipflipyong
Philip Yong
6 months
Out here at @aiDotEngineer. Thought that ā€œevalsā€ were some fancy tool, now I know it is short for ā€œevaluationsā€. It is to evaluate the performance of AI responses and s lot of times that’s done by using AI as a judge. Pretty cool learnings and validated thoughts I had on how to
0
0
2