@Tee_Gee_crypto merci ๐ right now it's relaisยฐ: it plugs your WhatsApp and LinkedIn into your AI assistant (Claude, ChatGPT, Grok). nothing US-specific, it's for anyone who basically runs their work from their DMs
sold my first company in 2022 (sign language translation for deaf people)
then traveled around the world
then AI video translation, pivoted to real-time conversational avatars for e-learning
now building new stuff from paris, aiming at the US. and building my own tools on the side because i can't help it lol
if you're a founder on a similar path, say hi. happy to connect
@hirahaniff stealing the "you asked for this" message ๏ฟฝ๏ฟฝ๏ฟฝ and the people who described the pain in their own words are also the best judges of whether you actually fixed it
"talk to customers before you build" is only half true now
a working product takes a week. so my loop to test an idea:
week 1 think
week 2 ship something that actually runs
week 3 try to convince first customers
under a month to know. vision, design and the experience matter almost as much as the calls now
@Matt_ships both, since it reads the whole chat, not just my side. I literally ask "what did people promise me this week that I still haven't received?" and it pulls out the "I'll send it Monday"s. then it drafts the nudge, I ok it, it goes out from my account
Most founders don't have a follow-up problem.
They have a "where did they say that" problem.
The price, the date, the "ok let's do it"โฆ
all buried in a WhatsApp chat from 3 weeks ago.
So this morning I asked my assistant:
"what did I promise people this week, and who's still waiting on me?"
It found 4. Drafted the replies. I said ok, they went out from my own WhatsApp.
@fezilemapela automated DMs are the tell. when a tool sends 200 messages while you sleep, the prospect can feel it. I still use AI on LinkedIn, just the other way round: it reads my inbox, tells me who actually wrote and what they want, drafts a reply, nothing goes out until I've read it
@Timmypaul2007 three business days is optimistic ๐ญ I got tired of it and plugged my WhatsApp into Claude. now every evening I ask "who's still waiting on a reply from me?", it lists them and drafts the answers, I ok each one. made it a little tool โ https://t.co/ZuLyNfVtER
@nickvasilescu the scene that sold me in Her wasn't the voice, it was Samantha sorting his inbox and telling him what mattered. that's the bit I built first: my AI reads my WhatsApp, tells me who's waiting on me, drafts the reply, I say ok, it sends. no earbuds yet ๏ฟฝ๏ฟฝ
it's relaisยฐ, the LinkedIn side. works in Claude, ChatGPT or Grok: add the connector, click Connect, sign in to LinkedIn on a separate page (it never sees your password). built it for myself โ https://t.co/3AsudLQjpm
LinkedIn is the last inbox founders still read themselves.
No assistant, so what matters sinks under the pitches.
My Sunday fix, one question to my AI:
"who wrote to me on LinkedIn this week, and what do they want?"
It drafts replies. I ok each one, it sends from my account.
@ZenithcoreAi Reaper with email access is brave. mine has my WhatsApp, and the first thing it did was list everyone who's been waiting on me. less scary than a debt number, same "oh no" feeling ๐ now it drafts, I say ok, it sends from my own account
@DimpleAroro23s personal side project that got out of hand ๐ relaisยฐ: plugs your LinkedIn into Claude, ChatGPT or Grok. "who wrote to me this week and what do they want?" โ it answers, drafts the replies, sends them one at a time after my ok. https://t.co/3AsudLQjpm
@0xKiyoro code mode for the reads is a no-brainer. the one thing I'd keep out of it: anything that sends. my WhatsApp MCP can search and summarize all day inside a script, but "reply to the client" stays a single visible call I ok first. cheap tokens, zero surprise messages
@zaimiri Chief of Staff as the only interface is the right call. the bot I'd add as #11: the inbox one. reads my WhatsApp + LinkedIn DMs, tells the Chief of Staff who's actually waiting on me, drafts the replies, nothing goes out until I say yes
@fordsmith that's exactly how it felt from the inside. each one taught me the same lesson, louder: build for a problem you hit yourself every day. that's how the WhatsApp + LinkedIn tools started ๐
@markshaw@claudeai love this. for me half the "names, numbers, dates" were already sitting in WhatsApp chats, so I plugged WhatsApp into Claude. "what's the address Paul sent me last month?" now takes 5 seconds. does yours pull from your messages too, or only what you voice in?
@knowRowan the approval step is the feature, not the limitation. my assistant drafts my WhatsApp replies to clients, nothing goes out until I say ok. an "autonomous workforce" sounds great until it confirms the wrong date to your best client ๐
@beamnxw finding them is step one. the part I kept dropping was step two: the few people who actually replied on LinkedIn, buried between recruiters and pitches. now I ask Claude "who answered this week and what do they want?", it drafts each reply, I ok them one by one
github issues as the coordination layer makes total sense
10 agents in parallel with no written way to decide = 1000 conflicts and zero progress. the issue tracker ends up being the boss
a boss is not a title, it's a document
A high school kid is recompiling the GTA6 to work on PC.
Whatโs more interesting is how theyโre github issues to coordinate tasks across agent swarms.
A new model for open source?