AI x UXR, MR & CXR
Earth
Joined December 2011
- Tweets691
- Following323
- Followers182
- Likes322
Pinned Tweet
Excited to share this new experiment by @erictli that helps you get fast usability feedback.
We built a tool that let's you test designs in minutes without needing to scheduling sessions or recruit participants.
Here's how it works: 👇
1. Upload your figma prototype and choose the screens you want to test.
2. An AI persona walks through it like a real user
3. You get a report with feedback and recommendations.
This is something between a heuristic evaluation and a usability test. It's not a replacement for testing with users but it's a quick way to get a second pair of eyes and catch issues before doing more research.
Comment below or DM me if you'd like to try it out!
Versive now supports SSO through Okta, Microsoft Entra ID, Google Workspace, or any SAML 2.0 provider.
Anyone your IT team assigns signs in with their work email and joins your org automatically with the default role, no invites needed.
Remove them in your identity provider and they can't get back in.
Enterprise plan only.
You can now open the Versive Assistant on a study and say "recruit 50 adults in the US who work full time on Prolific, $8 for 10 minutes."
It checks how many people match, shows you the cost, and drafts the panel. You still review it and hit Pay and publish yourself, no agent can charge your card.
Number questions in Versive can now be checked against earlier answers.
If someone says they see 40 patients a week, a later question about how many started treatment this week can be capped at 40.
The participant sees it as a hint, and if they change the first answer the check moves with it.
Also shipped in the last two weeks:
Matrix Allocation (one grid, up to eight scenarios, each column with its own total)
Study context for the AI interviewer
Memory is out of beta with focus and rules
AI tests that run about 3x faster.
Full notes: getversive.com/blog/october-…
Someone will always pick "None of the above" along with two other options. 🤦♂️
In Versive you can now mark an option as exclusive: picking it clears everything else, and picking anything else clears it. Works with max selection limits and "Other" text fields.
More below.
You can now set quotas per link or per answer option in Versive.
Need 50 people on iPhone and 50 on Android? Add a quota to the screener question and once one side is full anyone who picks it gets screened out right there while the other side keeps going.
Or do it by link. Send ?group=panel to a panel and ?group=customers to your own list, give each one its own cap, and when one fills up that link closes and the rest stay open.
Only completed interviews count, and you can set all of this up from Claude or ChatGPT through our MCP server too.
Versive is now in ChatGPT.
Go to plugins, search for Versive, sign in, and you can start your first study from the chat.
You talk through what you want to learn, it writes the questions with you, and when they look right it builds the study in your Versive workspace. Once responses come in you can ask it what people said about onboarding or where they dropped off and it answers from the transcripts, then it writes the doc or the slides for whoever asked.
It uses your Versive login so it only sees what you already have access to.
link: chatgpt.com/plugins/plugin_a…
crro retweeted
Across our benchmarks, the model sets a new standard.
It scores 52.6% on Terminal-Bench-Science 0.1, more than double Fable 5. On Terminal-Bench 4.0, it scores 55.8% against 42.0% for Fable 5.
ALT Benchmark table comparing Claude Fable 5.1 with Fable 5, Opus 5, and GPT-5.6 Sol across seven evaluations. Fable 5.1 leads on every row, including 52.6% on Terminal-Bench-Science 0.1 and 55.8% on Terminal-Bench 4.0.
We just shipped Versive's MCP server!
You can now connect it to Claude or ChatGPT and have them create your studies, pull your results, and build you slides or 1-pagers from the data that was collected.
It uses your Versive permissions so your agent can only see and do what you can inside the app.
Works with Claude, Cursor and ChatGPT
You over-ordered at dinner and you have to choose whether you eat the pizza or the pasta at the restaurant and take the other one home.
What do you pick?
The right answer is OBVIOUSLY to eat the pizza at the restaurant and take the pasta home. Most of my friends and everyone at @getversive disagreed. @erictli suggested I run a study with 100 people.
Results below 👇
