Michael Wang (@wangmander) header
Michael Wang (@wangmander) profile picture

Product Design Leader · SF Building https://teapawt.net/t.co/WCK7PrZ4Tx

Joined October 2008
Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)Photo by Michael Wang (@wangmander)
can @AnthropicAI or @OpenAI replace phones already. done with @Apple
3
128
I'm 4.8 MAX'ing EVERYTHING, starting with the important questions: is a hotdog a sandwich? let's goooo @claudeai
20
No account signup was the hardest constraint. Means no user state, no templates, no customization flows. But it forced better defaults. When you can't ask users to configure anything, the first output has to work.
1
13
Everyone is reading Cisco backwards. The scary part is not "nearly 4,000 jobs cut." The scary part is: stock hits a record AI demand surges company cuts anyway Markets are starting to reward companies for proving they can grow with fewer people. That is the new job market.
1
1
50
This is why I'm building Guildy. Applying is easy. Landing is hard. Manage the round, not the application. Track the pipeline. Prep by stage. Compare salary and opportunity. Stop running interviews from tabs, notes, and memory. guildy.ai
1
37
5 companies. 1 week. ~2,700 jobs cut. Freshworks. Coinbase. Cloudflare. DeepL. Upwork. Different businesses. Same new pattern.
1
1
127
For candidates, the lesson is simple: applying is easy. landing is hard. The bottleneck is no longer finding roles. It's staying organized, prepping by stage, knowing your risks, showing up sharper than the 100k+ others in the market.
1
42
Breaking down each company this week + building around the new reality. a command center for the interview process. guildy.ai
32
Different models for different jobs. GPT writes the product spec. Claude builds the actual components. Gemini reviews the full codebase for coherence. Each one is better at its specific thing than any single model handling everything.
61
The multi-model split is real now. GPT handles strategy. Claude builds specs and code. Gemini reviews the codebase. Most debugging is figuring out which model introduced the problem, not finding the bug itself.
1
75
A senior engineer runs $150-200/hr. A two-day sprint scoping theme systems and dependency audits would cost $2,400-3,200. This week it was one person and three AI models. capsuleweb.site/?utm_source=…
1
59
Working across 3 products this week. CapsuleWeb — $1 instant website generator, live RawIQ — $1 adaptive IQ test, almost live Guildy — automated job tracker, private beta 25-30hrs sr eng work. Good hard work @claudeai my boy.
57
GPT is good at planning what to build. Claude is better at actually building it. Gemini catches things both missed during review. Took a while to stop treating them as interchangeable.
47
Nice fit for a cafe that just needs hours, a menu, and a location link on one page. No CMS, no monthly fee. One prompt, one dollar, live in 30 seconds.
60
We had 1,000 random outputs and users didn't care about any of them. Then we grouped everything into just 6 theme families — and suddenly the content felt intentional. Less randomness, more resonance. full breakdown in the replies
1
71
Broke a 586-line renderer into 11 components last week. Not because it was failing — because I couldn't reason about theme logic in a single file anymore. Smaller files, faster decisions. The refactor paid for itself the same afternoon.
63
Submitted CapsuleWeb to Hacker News if anyone wants to check it out or weigh in on the no-account tradeoff.
1
78