Valentin

Early August, @_valentin and I spent 4 hours arguing about the future of @ToltHQ. I wanted a monorepo, Hono and proper end-to-end types. He wanted to keep doing whatever we were doing, which as far as I could tell was vibes and any. Half a workday. At least 30 jokes about the other one’s code. Eventually he gave in, mostly so I’d stop talking. Since then we moved our whole API into a monorepo, shipped TypeScript and Python SDKs, packages for embeds, tracking and webhooks, an MCP server, and brand new docs. The team has never shipped this fast, and we’ve started on a new portal too (more on that soon). Valentin, I’m available for another fight whenever you are.

introducing @loopsterGTM - one GTM platform for a flat price, where the data is free. we’re putting agencies back at front seat by flipping the pricing upside down. all the data is given for free. it’s sourced by the users, and it’s given for free to the users. here’s how it works /n

About 90% of @heyreach_io's revenue was inbound when we hit $10M ARR, we hadn't run a single ad. This was possible because we ran word of mouth like a channel we could control, with someone on the team owning it every day. Here are 10 underrated things we did to get there: 1/ We designed our partner program so creators earn more when their customers stay. → 50 creators and 16 agencies are in it, and none of them get paid for the post itself. → They get a share of what their referrals pay us. Creators get 35% on our Growth plan for the first year, when the usual B2B SaaS commission is around 20%. → If a customer refunds or downgrades, the commission comes back, so there's no point sending us people who'll cancel in month 2. → That partner and affiliate program has brought in over $3.1M so far. 2/ Every mention gets a reply from a real person. → Every time someone mentions HeyReach on LinkedIn, 1 of us comments on the specific thing they wrote. → 1 person owns this every day, and a small account gets as good a reply as a big one. → In our data, people whose mentions we reply to are about 3x more likely to mention us again. 3/ Every mention also goes to sales. → We log every mention: who posted, which segment they're in, and what made them post. → We send that list to Nadja, our Head of Sales. Her team reaches out while the post is still fresh, and the first message opens with what the person wrote. → They posted about us first, so it's never a cold pitch. 4/ 10 of us post under our own names, with money on the line. → 10 people on our team build their own audiences, with opinions I don't approve first. → We built a quarterly posting competition for it. You opt in and commit to 1 to 5 posts a week. Hit your number and you get a bonus. Miss a post and you pay out of your own salary. (I hate losing, even at ping-pong.) 5/ We publish our users' playbooks as our case studies. → We call them Outbound Outliers. We take a real user's campaign, rip it apart, and publish the whole playbook: the steps they ran and the results they got. → The user gets the credit, so they share it with their own audience. 6/ We gave webinar attendees a reason to post their results. → Attendees run a campaign, post their results on LinkedIn, and the best ones get rewards. → Their audience sees HeyReach working for someone they already follow. 7/ We treat engagement on our team's posts as a buying signal. → We qualify whoever engages with posts from our team's active LinkedIn profiles. → We stack that with other signals: a site visit, trial activity, ad engagement. → The warm ones go to sales while the signal is fresh, and the message opens with whatever the person did. 8/ We fixed the product before we pushed more traffic. → Before anything else, I ask: if traffic doubled tomorrow, would the business be better, or just busier? → If it's busier, we fix what new users see first. We tier every signup against our ideal customer, and the best fits get a human on day 1. 9/ We waited until $10M ARR to run any ads. → First 5 months of LinkedIn ads: $21,404 spent, 24.9x ROAS. → The same ad made back 5.7x on a broad audience and up to 91x on people who already knew us. I don't fully trust that 91x, because some of them would have bought anyway. → When a post from our team takes off with $0 behind it, we test ad budget behind that exact post. 10/ We stopped trusting the attribution report for word of mouth. → Picture a buyer who heard about us in a Slack thread, saw 3 posts from our team, then clicked a Google ad. Google gets the credit. → So we check 2 other things. The first is what share of revenue comes in without anyone reaching out first. That's the 90% from the top. → The second is whether word of mouth keeps running while I'm on holiday.

introducing Tolt Agent: the AI affiliate manager. open it from any page and ask anything in plain English. it knows where you are and answers from your actual Tolt data in seconds. it can spot things you might miss, like fraud signals, unpaid commissions, or partners going quiet. and it doesn’t just answer questions. tell it what to do: approve partners, adjust commissions, record customers, and more. go try it. btw what do we name him?

These past days I've been rehauling @ToltHQ's MCP and built our own agent 🤖 Here it is building a flow for me. Maybe in this day and age UI is slowly less used as MCP tools start to become more powerful...