Partner Story ยท in conversation with Peter Morris, developer and co-founder of Ortivo. This is a written interview submitted through Submit Your Story. The claims below are the founder's own, presented as they were given to us. No payment was involved.
Peter Morris and Thanasin Thanasarnchokephiboon, husband and wife and business partners behind Ortivo, in Sydney. Photo supplied by Peter Morris.
Peter Morris used to start his working day with the same routine. Type a question into one AI chat, copy it, paste it into a second tab, then a third. Compare the answers by eye. Pay for three pro subscriptions, $60 to $80 a month in total, and still run into a rate limit on the one afternoon he needed to get through ten documents in a row.
Ortivo is the tool he built to end the tab ping-pong. Morris is a software developer and consultant in Sydney. He builds the product together with his wife, Thanasin Thanasarnchokephiboon, a data analyst with a master's degree in industrial engineering, who has also studied healthcare and data science at the University of Sydney. Ortivo is owned and operated by Deptly, her sole-trader business, and the two work on it as partners.
One question, several models, one bill you control
The idea starts from something most heavy AI users notice sooner or later: different models are good at different things. Some are quick at broad synthesis, some are strong on long documents and code, some are good with live web results, and open-weight models through OpenRouter can be much cheaper.
In Ortivo you type a question once and send it to up to four models at the same time: OpenAI, Anthropic's Claude, Google's Gemini and text models available through OpenRouter. The answers stream in side by side, each with its own progress.
The money side is built around "bring your own key." Users connect their own API keys and pay the AI providers directly for what they use. According to the how-it-works page, the current pilot has no workspace membership charge. A cost estimate shows the expected input and output before you send, with prices synced every hour from the OpenRouter and LiteLLM catalogues. Ortivo also makes no automatic retries, so a failed request does not quietly double the bill, and it caps use at 100 provider requests per account per day.
For people who want to look things up without any AI key at all, there is built-in web and news search through Brave, paid for by Ortivo within daily limits.
The council that can't copy
The feature Morris is proudest of is the AI Council, and in particular its Live Evidence mode.
It works in three steps. First, Ortivo pulls up to five current web search snippets and shows them to the user, with links and the time they were retrieved. Then two or three reviewer models each assess that same packet of sources, and none of them sees what the others wrote. Finally, a synthesizer model reads the sources and all the finished reviews and sorts the result into agreements, disagreements, supported findings and claims that remain unresolved.
Every stage shows a preview of exactly what will be shared with which provider, and waits for the user to approve it.
The lesson: agreement is not verification
The blind design came out of a failure. In the first prototype of the Council, Morris let the second and third models read the first model's answer, so they could "debate" it.
He tested it on a tricky question about cloud pricing. The first model quoted an outdated 2024 price tier, with complete confidence. Because the answer sounded authoritative, both follow-up models agreed with it and built their own calculations on top of the wrong number.
"That failure was the lightbulb moment," Morris says. "Agreement between models that can see each other's answers is not verification. It is just an echo chamber."
So he took away the cross-model visibility. Each reviewer now had to judge the same five live search snippets on its own. "The disagreement surfaced immediately, and the synthesizer caught the outdated figure," he says. "Building for human verification rather than autonomous theater changed the entire product."
The product still carries that lesson in its own wording: the site reminds users that agreement between models does not by itself prove a claim is correct, and that important facts are worth checking.
Research with a human in the loop
Two more tools follow the same logic of "show first, then spend."
The Research Agent runs structured web research in three fixed sizes: Quick uses two AI requests, Standard four, and Deep five. After the planning step it stops, so the user can approve or edit the research questions before the remaining paid requests go out.
The Document Reader takes PDF, Word, Excel, CSV and Markdown files, up to three at a time. It can also rebuild common charts from Word and Excel files as images, so a vision-capable model can look at the actual chart rather than a description of it.
Around this sit a few practical extras: an encrypted decision notebook that is never sent to an AI unless the user attaches a note, conversation history that is off by default, scheduled private research briefings, and a library of 2,024 starter prompt templates across 112 categories.
Where it stands today
Ortivo launched on 15 October 2026, and Morris is straightforward about the stage it is at. The pilot has a small but growing number of registered users, with new sign-ups arriving every day. So far they are mostly independent software engineers, management consultants and academic researchers, who use it to replace several $20-a-month chat subscriptions with pay-as-you-go access to several models at once.
What comes next
Next on the roadmap are shared prompt workspaces for teams, side-by-side difference views for comparing code and legal clauses, and more presets for open-weight models through OpenRouter.
Morris and Thanasarnchokephiboon are inviting developers, researchers, analysts and small teams to try the workspace free with their own API keys and tell them about the edge cases they find. Ortivo is live at ortivo.com.
---
This is a Partner Story: a written interview with a founder building with AI, submitted through Submit Your Story and published free of charge. Statements about the company's product, users and pricing are its own. Building something with AI? Tell us about it.
Would you try this product?
