Eric Khun

Buffer has launched dark mode for halloween 🎃 👻 . How cool is that !? Love the 1 click to try with the logo 🌞 / 🌚 , such a nice touch coupling context and feature on-boarding

Yongkang Street: 7-Eleven's Only by @Stacy - Convenience Store Marathon Explore Yongkang Street: 7-Eleven's Only, a 4-store convenience chain covering 0.5km in Taiwan by @Stacy.

Yongkang Street is one of my favorite area in Taipei. For some reason, only 7/11's are on this street? craftmygame.com/game/conveni...

Area you team Dark Mode? It's now available at Buffer ! 🌚 / 🌞

How about a matteress connected to @pagerduty ? 🫣 x.com/m_francesche...

With AI, the cost of creating things has never been that low.

Here's a look at the main services powering Buffer.
Adding a service is easy. But it comes with the burden of maintenance that bites you over the years: security patches, updates, keeping context. One of our goal: keep the stack dead boring. Boring = manageable and sustainable.

AWS cascade failure:
- DNS issues → Couldn't resolve DynamoDB endpoints
- DynamoDB fails → EC2's internal systems depend on DynamoDB, so...
- EC2 launch failures → Network Load Balancers couldn't launch properly, so.. broke network for... Lambda, CloudWatch, SQS, RDS etc...

I'm always half-joking that we should make a Taiwan convenience store game: longest chain of 7/11 or Family mart where you can always see the next one from where you're standing. 10km+ has to be possible, right?

Half-joking that we should make a Convenience Taiwan game: longest chain of convenience stores where you can always see the next one from where you're standing. 10km+ has to be possible, right?

Claude code just gave me ideas I didn't even think of. How good of a PM can it be?

The UI/UX in Claude Code is incredible - it dynamically reshapes itself based on the context of your questions

claude code now has reverse backsearch (ctrl+r) . Don't ned to retype your entire prompt every single time : D

Quick sneak peek on the craftmygame game generator and editor we've been cooking:

Just looked at teammates tenure at @buffer:

We're 70:
- 47 have been here 4+ years (67%)
- 37 have been here 6+ years (53%)

I believe remote work + 4-day weeks means people don't burn out and leave. And we're more productive than ever and the business keeps growing.

Super nice AI integration in @notionhq on making complicated query easy to write.

really cool interactive graph on the cause mortality according age and race
flowingdata.com/mortality/ . Data for 2005-2014 from the CDC

Cloudflare AI workers allows you to query many open source LLMs with a generous free tier that reset daily! super nice if you start a project that need some simple AI features

can now create a unique (and playable) character for any game !

For LLM and design: a good pattern I've found was asking you LLM to give multiple versions of the same thing and you just need to pick one at the end

No more guessing when you'll be about to reach Claude Code limit with /status !

Now, when you start Claude Code 2.0 , now they allocated out of the box 45k tokens (22.5%) for [autocompact + output tokens] . Could this improve the bad performance when it starts auto-compacting?

Plan Mode with Claude Code 2.0 is still available with Shift-Tab shortcut. Why did they ditch it in the /model? I still do believe knowing what is the intended plan and brainstorming with the LLM before it tackles anything is how we should use them

Did Claude Code just removed the Opus Plan mode? 🤔

App stores are about to get disrupted and in a fun way. Excited to see all the new stuffs with original way to do things. Only developers or companies with $$$ could create apps. Now a new type of creative people are about to enter the game

Codex CLI finally added `/limits` commands to know how far/close is your usage !

maybe that was too much to ask

I probably had too much expectation for the codex high model

Codex finished building the app in just 3 minutes?! Did that really work? 🤔🤔 I don’t even know how to try the new app. Looks like I need to download Xcode… and that’s gonna take 12 minutes—4x longer than Codex took to build it! 😂

I've just started completely new mac app project with openai's codex high. No skeleton he has total freedom and start from 0 file. will this work? Trying this tips where it tracks its work in a markdown file until completion x.com/embirico/sta...

OpenAI just released a new toy, chatters saying it's better than claude? 🤩
LMarena benchmarks was putting chatgpt at the worst at coding. Maybe codex will change it?

benchmark here: aidailycheck.com