Ben Cheng

Actually I started with sprites.dev first, but it's public url proxy intercepted .well-known/ for MCP auth, which is the path I need...

shellbox.dev is probably the best instant server option for vibe coder like me who just want to host some random vibe code things cheap and no BS.

@flydotio Love the idea of sprites! It would be great if the cli works on Android termux, so I can give up my ssh box and start vibe coding on Sprites :)

@kepano 1. Generate daily digest from sources 2. Plan my day, with tasks from inbox, calendar and make it into a schedule in daily notes, including empty time between meetings for tasks 3. Generate LinkedIn posts from notes changed last 14 days

if you're using Obsidian with Claude Code, tell me about your workflow, and what you've used it for

Planning to add more agents for daily/weekly reviews and vault maintenance tasks. For Obsidian users I strongly recommend try to combine it with Claude Code.

The experience of creating "commands" and "agents" in Claude Code reminds me of that old joke: professors use the ultimate programming language - natural language - they just tell the grad student what they want, and someone else figures out the implementation.

Spent the afternoon using Claude Code to organize my Obsidian vault. Wondering why I haven't thought of using Claude Code this way earlier. LOL What I tried today:

And don't deprecate models before stable replacements exist. Current alternatives to 2.0? If not 2.5, Gemini 3 Flash and Pro are both in PREVIEW—not production ready. Wait until 3.0 Flash Lite hits GA before retiring 2.0!

More over, GCP need to provide viable alternatives. Recommending an upgrade to 2.5-flash when it shuts down just 3 months after 2.0 (June 2026!) is ridiculous.

Unpopular opinion: @googlecloud AI model retirement strategy is disappointing. While I love Gemini personally, production AI agents need stability. Model changes often break workflows and reduce accuracy.

After 5+ years with @SuperhumanMail, I'm moving on. They kept adding features nobody asked for while breaking their core promise: being the fastest email client. Still no keyboard shortcuts on Android, can't even copy email links on mobile. Sometimes less is more.

I think the launch of Sora indicates how desperate OpenAI is. The goal was to create an AGI (or at least an AI agent that replaces most of the jobs) which is profitable enough for Sam Altan to provide UBI, instead it's launching another social network (and a bunch of tools)

Ok I have to admit this is so cool.

Postgres 18 has been released, with Async I/O support. Previously, all read requests were blocking, but with this update, they are no longer, delivering massive performance gains for read-heavy applications! It's enabled by default on Postgres 18!

They discussed achieving PMF in US market and shared specific tactics for promoting on Product Hunt, Reddit, Twitter, and Threads. ​The craziest part? The entire event was in Mandarin, yet I feel I learned more about US go-to-market than I have from most English materials.

I never realized how many AI products are coming out of China until today. I was at a half-day event in Shenzhen, featuring founders from Trickle, Gamma, Fellou, Kuse.ai, and more—all China-based teams with a go-to-market strategy for the US.