Chris Wong (@zeviance) header
Chris Wong (@zeviance) profile picture

Entrepreneur wannabe.

Joined June 2007
Photo by Chris Wong (@zeviance)Photo by Chris Wong (@zeviance)Photo by Chris Wong (@zeviance)Photo by Chris Wong (@zeviance)Photo by Chris Wong (@zeviance)
psa. if she's into you, she's probably a chinese spy.
20
Amazon Photos has one of the dumbest backup performance bugs I’ve run into. It said it was “uploading” my GoPro videos, but Task Manager showed 0 network traffic while my external HDD was pinned. So I started digging.
1
186
Turns out “Upload started” doesn’t mean upload started. Amazon first reads the entire file to calculate an MD5, then checks its local DB. Most of mine ended with: MD5 matches ... Skip creating. So GBs read, 0 bytes uploaded.
1
17
Google's biggest AI-mistake - dropping Gemini CLI support in favour of Antigravity CLI. 1. Antigravity is not ready for prime time. 2. That's underestimating the friction to switch providers. 3. Most of the core features from Gemini CLI were not supported. 🧵
1
69
Something as simple as supporting folders that has a dot in the file path. 4. Gemini CLI was open-sourced and backed by active contributors that were making it better everyday. Good luck funding another open-source project going forward.
17
Putting on my AI bro tinfoil hat for one sec.. We're now speed-running early 2000 internet bandwidth issues, but with LLM. Hear me out - - we're simultaneously running into speed issues, availability issues, quality issues, accessibility issues, tier-metered issues.
1
13
This is textbook consumer 56k modem vs comcast business class enterprise grade ethernet internet in early 2000s. I am still not 100% convinced off-grid LLM (aka mac mini) is the answer, but as a hedge.. not a bad idea.
21
Don't sleep on this. Qwen3.6 is getting seriously close to Claude Opus 4.6, and arguably better in terms of consistency and service availability. Forget about the leaks, forget about the features smoke and mirrors, there's a new competition in town.
40
Quick recap. 1388 commits in March, average about 50 PRs per day. Switched later part of March to focus on building the "factory" instead of designing the "car". Gonna aim for broader factory dogfooding next week. TBD if I want to opensource it.
1
13
nts: Don't mistake motion for progress. Keep shipping. 🦾🦾🦾🦾🦾🦾
7
- Massive token burn because entire conversation gets rebuilt as cache_creation tokens instead of hitting cache_read github.com/Rangizingo/cc-cac… I guess sometimes opensource is actually good for everyone.
1
15
Who knew the key to unlock development velocity was a Claude Max subscription. 923 commits in 2 weeks, unreal. Excited to see how far we can go when I have the dev loop architecture dialed in...
1
26
Dear Founder, I DMed you but you still ain't callin' I left my Calendly, my cell, and my Loom link at the bottom I sent two follow-ups back in autumn, you must not-a got 'em There probably was a problem with the spam filter or somethin'
1
1
1
2
621
Sometimes I script the subject lines too sloppy when I bot 'em But anyways, what's up, man? How’s the churn? My AI’s got the "agentic" workflows that you gotta learn I’m building a side-hustle too, I’m 'bout to be a founder If I get a Series A, I'ma call it "Bounder"
1
2
25
One of the weakest link of Generative AI is the lack of context around the subject; Case in point - ignore the fingers, ignore the face, ignore how the liquid looks like in the cup. Look at the artifact, zoom in and take a look at the coffee cup. 🧵 #generativeAI
1
18
2 weeks ago the same coffee shop posted a photo on their Instagram showing their to-go drinks packaging - it's paper brown. The one in hand is a green one. We're now in a very interesting timeline. IYKYK. #generativeAI
31
Special shoutout to @Cloudflare @sidchatterjee @allnoteson @craigsdennis @openclaw @argenistherose and other contributors that built the mountains and made this possible.
1
4
290
I didn’t build the infra, and I didn’t build the brain. I built the bridge. It wires together @Cloudflare’s Moltworker architecture with the ultra-lean Rust runtime from @ZeroClawLabs (<5MB RAM). The result? A 24/7 assistant that actually fits on a $5 Workers plan.
2
2
5
740
Stop buying $600 Mac Minis just to run an AI agent. Everyone wants an always-on @OpenClaw instance, but the memory footprint usually makes cloud hosting expensive or unstable on cheap tiers I just launched moltworker-zero Your private agent for ~$5/mo 🧵 github.com/zeviance/moltwork…
7
9
58
10,955
Still in disbelief with the assistant of AI coding tools, I managed to churn out a web platform in 2 days with 11k lines of typescript code. Gemini/Claude code really changed the game. Schema generation/lint/test cases baked in. I am not even a web developer.
37
Idea - bring the MUJI philosophy and apply it on day to day software. Simple, no brand name, no gimmicks, no ads, single purpose, good functionality, good feature, good perf. Internal software suite could be a good test bed to grow these internal systems.
29
Startup horror stories part 2. You know you’re in for a great time when they use users’ Social Security Number as the database unique primary keys, non-hashed. #doYourDueDiligenceKids
32
Startup horror stories part 1. You know you’re gonna have a great time when they offer 23” Dell Monitors for engineers to work from office. #doYourDueDiligenceKids
27
Hey @SouthwestAir looking great out there @flySFO
1
30