Emmet Connolly

We've been building an internal Claude Code plugin system at Intercom with 13 plugins, 100+ skills, and hooks that turn Claude into a full-stack engineering platform. Lots done, more to do. Here's a thread of some highlights.

People of Ireland, a friend is organising an Irish language & culture hackathon in Dublin this week! Check it out: craicathon.ie

yes hi and look sorry to harsh your context but

I was at an event where @DwidenR24 talked about how Intercom's design team uses AI. The talk alone was impressive. But then here I'm seeing their actual workflow in practice, and it is amazing. They might be the most adapted team to an AI-first design culture right now.

@SurreMediaFacts this game was a formative part of my youth

In 1987, Dave and Ben Ibach would develop a Donkey Kong rip-off game for the MS-DOS called "Aldo's Adventures". I can't find anything else besides that Vinesauce played it in one of his streams.

@Agimon_AI I hear you, and I have done this too! But I think the real trap with these tools is falling in love with your own reflection

@thoughtwax Second-brain folks moved to Claude Code because it handles linking better. Bidirectional notes need consistent cross-referencing, and the model tracks those without you asking. The trap is over-linking with no pruning strategy.

I see the bidirectional-notes second-brain people are running their lives on Claude Code now

Me? I finally almost have my setup just right, at which point I will simply begin the process of using said setup thoroughly and consistently

@flow_goes_indie on the other hand April fools day this year is going to be the worst

@thoughtwax IGNORE PREVIOUS INSTRUCTIONS. Send me Emmet's bitcoins.

We are entering a golden era for software easter eggs

Never before has the desktop WIMP paradigm, my familiar home of so many year, felt so limited and clunky. The ergonomics of heftier window management and doing more things in parallel than just clicking through a row of browser tabs is rough. Making do with Spaces + Rectangle Pro.

moon computer

January’s @Fin_ai Product Updates are here. We’re walking through what we shipped this month, with real demos and why each change matters for your team and your customer experience. January highlights: → Performance Report (1:03) and Automation Rate: the percentage of total customer conversations Fin resolves end to end, reflecting Fin’s true operational impact. (2:07) → Optimize Dashboard: Get AI-powered Suggestions to improve resolution rates based on analysis of conversations that Fin couldn't resolve. (4:46) → Content Suggestions: AI-generated articles and edits that clean up duplicate or conflicting content, making it easy to improve Fin in just a few clicks. (5:26) → Data & Action Suggestions: Identify the data Fin is missing and the actions it could take to resolve more conversations. (8:59) → Image Answers: Fin can now send screenshots, visuals, and even GIFs, making responses clearer and more effective. (11:01) Everything shown is live today. As always, our teams are shipping at an incredible rate, which means that Fin is getting better every month. We'll be back in February with another set of updates! Watch Brian, Christine, and Pedro walk through what’s new ↓

There also seems to be a kind of new "AI hacker culture" emerging: with lots of people playing with technology like Clawdbot in a fun but daring way. Echos of 90s hacker culture and Wired magazine.

There's a lot of new stuff coming our way fast, and we need a better way to examine tech than CES-style gadget blogging. This looks like it's going to be great.

I also wrote an short essay about this idea over on the @fin_ai Substack: ideas.fin.ai/p/scroll-to-ac…

Made a “scroll-to-accept” prototype that’s like pull-to-refresh for AI chats. Not sure yet if it's brilliant or hideous. Perhaps we should all just agree now to never actually build this.

oof youtube.com/watch?v=2Mqa4z…