We’re an applied research lab building models that reflect how reasoning, emotion, experience and social influence shape our decisions. Model what happens next.
Let Wren handle your QA workflow. Describe what your app should do in plain English and Wren's agents test every path, with browser evidence for each failure.
Fully offload your QA. Book a demo below.
We’re live on Product Hunt! Meet Wren, @seldonhq's behavioral QA agent.
Same goal, different users, different paths. Wren tests what happens when someone retries, goes back or changes their mind, and records what breaks.
Give it a feature you’re about to ship and watch it play out. Claim your first 10 runs free 👇
https://t.co/bUyaXo1wQ0
Seldon’s Wren, a behavioral QA agent that uses your product like customers do, by @CharmanTom and @HiggsOlivia has launched on Product Hunt! It would be great if you gave it a look.
https://t.co/XqdyvaO6LX
Seldon simulates how people use products to help teams spot friction.
Behavioral fidelity measures how closely those simulated choices match real users’ choices: where they hesitate, go back or leave.
https://t.co/yW7pfoSJC9
Halloween came early at the Seldon Speakeasy! Thanks to everyone who came through last night. A proper night. More soon check out our Luma for more events
@dair_ai This is the core problem with letting an assistant play the user. Real people aren’t cooperative or explicit! They skim, misread, hesitate, give up. Simulated users need their own goals, knowledge and patience or every flow looks easier than it is
@myfear Separate the builder from the judge. When the same agent writes the code and the tests, the tests mostly prove the code does what the agent intended. Independent checks from ICP's perspectives in plain English, run against the real app in a browser are the way to go
@addyosmani Agents reviewing agents is pretty much the only way this scales. But a clean diff doesn’t mean signup still works. Something still has to click through the real app like a user would, and that’s the part that falls over when you’re shipping this fast
@cursor_ai And testing the running app only works if the tester behaves like your customers. Not a scripted happy path - but personas built from how real users move! what they’re trying to do, where they hesitate, what they skip. Otherwise the flow passes QA and still fails the customer
@asad0801 This is the core problem. Pass/fail was built for deterministic software. Agents need behavioral evidence: what it actually did, in what order, and whether a real user would have ended up where they meant to.
Any stack. Wren, our QA agent, tests your real user flows.
3 prizes: best overall, most original, people’s choice. Then demos + drinks behind the red velvet curtain.
https://t.co/4ybKRDxSR1
Ship Happens: a Seldon hackathon
Nov 7 · SoMa, SF
1 subscription you’re sick of paying for.
1 day to build the replacement.
1 of us has to use it live at demo time.
Up to $5k + Claude credits with support from @AnthropicAI
Sign up link below
On September 4th we launch /BLOK.SIM/01 - part party, part experiment, part glimpse into the future. 150+ AI minds dropped into a simulation: neon corridors, interrogation rooms, UV poker.
Black Mirror IRL 👀
https://t.co/KQas1Z7C7O