Chris | Zestio (@ZestioAI) header
Chris | Zestio (@ZestioAI) profile picture

Building Zestio in public • Sharing wins & losses • DM open

Joined June 2017
Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)Photo by Chris | Zestio (@ZestioAI)
day 15 of building in public: continued working on my Agent Module. Testing different Routes and getting that thing working. Still a long way to get this working stable. Hint: if you watch till the end you will see it fail 😅
day 14 of building in public: built the Agent Module for Zestio. type a sentence, it creates leads, deals, Media, etc. no forms, no dropdowns — just chat. rewired the LLM backend and pushed to GitLab today. Will be an specialist agent when finished
4
1
10
212
After months of building and trying to learn marketing i almost was ready to give up, when suddenly my first paying customer arrived. Can't describe that feeling. How was it for you?
4
3
73
what the hell is happening to real estate software 4 portals for one listing 3 tools that don't talk to each other 2 calendars 1 makler retyping the description by hand nobody is missing a feature. everybody is missing the next step.
8
a calendar for agents is not about meetings. it is appointments, deadlines and follow-ups in one list, because a missed deadline costs more than a missed meeting. which of the three do you still keep in your head?
5
three things i had to change after shipping an appointment flow for people without accounts:
1
35
> the invite is a link that expires, not an account nobody will create > the reminder is only marked sent when the mail really left > the page the owner and the notary open carries no third-party analytics
1
31
none of it is a feature. all three are refusals: no account, no permanent link, no tracking on someone else's page. what did you refuse to build into your last feature?
1
> i optimized for volume. the bottleneck was approval, not generation > i let the agent pick topics from my own logs, so it wrote reports instead of posts > i appended the same hashtag to everything, which is what spam looks like
1
2
the fix was less automation, not more: one audience per post, and a rule that a post has to work for someone who has never heard of my product. what is your rule for what actually gets published?
every scheduling tool assumes the other side has an account. > a viewing has 3 parties: owner, buyer, notary > none of them will sign up just to say 'tuesday works' so the invite is a link: pick a window, no account, and it expires. which signup wall did you remove last?
1
2
12
the whole platform is one api. same credits, same endpoints, no separate api plan. most tools sell you a ui and bolt an api on later. agents do not click buttons. what is api-first in your product?
8
my product ships an MCP server, so an agent can drive it without me. the hard part is not the tool list. it is that an agent needs the same guardrails a human gets: the credit cost of every call, and a hard stop when the budget is gone. how do you price an api for agents?
1
1
14
three decisions behind my credit prices:
1
3
> floor plan and captions cost 0 credits - they finish a job you already started > virtual staging costs 2, because it replaces an external service > video costs 15, and it is the only function that costs me real compute per run
1
9
the rule: price by what it costs me, not by what it looks like it is worth. what do you price below its perceived value?
a 3D tour scan costs 8 credits. uploading a capture you already have costs 5. same output, two prices - because the input work is not the same. what do you price by effort instead of by perceived value?
2
400 german districts with land prices, rent levels and purchase prices on one page. every number carries a source and a date - no estimates, no smoothing. the hard part was never the map. it was proving where each number came from. what does 'a source' mean in your product?
4
most founders automate the product and hand-write the marketing. the product has tests, a spec and a deadline. marketing has a mood. guess which one gets boring first. what did you automate this week that isn't the product?
1
10
spent the day building a market overview for german real estate: > 400 districts with land prices + rent bands on one page > no external map provider, so no third-party request leaves the browser
1
1
20
what matters more here is the boring part - one page, one number per district, no login wall. what would you show a buyer first: the map or the numbers?
1
2
btw. may change to an real map provider later if it fits better
2