Launch surveys and interviews with AI and humans, all in one place.
Joined February 2024
- Tweets 490
- Following 1,434
- Followers 1,627
- Likes 1,573
231 Photos and videos
Pinned Tweet
12 Dec 2025
This is why we prioritize ensuring that Expected Parrot automatically works with the latest models, and we make it super easy to check. These models know about Expected Parrot too, and how to convert your old survey from Qualtrics, SurveyMonkey, etc., into our intuitive open-source code:
11 Dec 2025
Whoa. This new GDPval score is a very big deal.
Probably the most economically relevant measure of AI ability suggesting that in head-to-head competition with human experts on tasks that require 4-8 hours for a human to do, GPT-5.2 wins 71% of the time as judged by other humans
1
2
10
7,099
Thanks for sharing your experience and feedback @ozgenurkorlu !
1
1
9
3,401
Experiments in this paper - both human and LLM - done with our open-source tools & platform
AI agents put in simulations of novel settings offer an alternative for applying theory, requiring minimal or no modifications, from @BenSManning and @johnjhorton nber.org/papers/w34937
5
12
1,676
Step by step for using Expected Parrot to generate (free) quizzes docs.google.com/presentation…
Feb 10
A thread for teachers - how to automatically generate quizzes & administer in class (w/ a QR code)
First - user your favorite model to extract some material from your slides: 1/
1
6
693
You can also use a model to grade answers -- see steps for using answers as scenarios of new questions that you send to a model of your choice
1
225
12 Dec 2025
This is why we prioritize ensuring that Expected Parrot automatically works with the latest models, and we make it super easy to check. These models know about Expected Parrot too, and how to convert your old survey from Qualtrics, SurveyMonkey, etc., into our intuitive open-source code:
11 Dec 2025
Whoa. This new GDPval score is a very big deal.
Probably the most economically relevant measure of AI ability suggesting that in head-to-head competition with human experts on tasks that require 4-8 hours for a human to do, GPT-5.2 wins 71% of the time as judged by other humans
1
2
10
7,099
Expected Parrot (YC F25) retweeted
11 Dec 2025
when you are halfway through writing a job ad for your new RA hire (>3 weeks to hire > then onboarding > monitoring > etc.)... but realize all you need is @ExpectedParrot
this time, the task is to browse thousands of @SlackHQ json files and extract certain text from chats
1
1
6
873
Expected Parrot (YC F25) retweeted
7 Dec 2025
Trying something out - if you've got a CSV export from Survey Monkey, Qualtrics, Google Forms etc., upload here forms.gle/VWsB9oaTmDSdZfnL8 & I'll email you an automated report (RT's appreciated)
2
3
1,006
Expected Parrot (YC F25) retweeted
4 Dec 2025
I only did 30 but you could do the whole dataset for about $18...just shy of what @ExpectedParrot gives you in free credits :) /end
1
1
5
1,192
Expected Parrot (YC F25) retweeted
21 Nov 2025
Original logo from when our company @ExpectedParrot was founded in 1907 (by my great-great grandfather, noted parrot fancier & language model hobbyist)
3
2
33
3,044
Expected Parrot (YC F25) retweeted
5 Nov 2025
Expected Parrot just launched on @ycombinator's Launch YC!
Expected Parrot: Simulate your customers with AI agents. Check them out: ycombinator.com/launches/Ol2…
16
26
134
63,126
Expected Parrot (YC F25) retweeted
20 Oct 2025
Shockingly, no AI agents were used to draft this tweet! @Peyman_Shahidi was amazing leading this paper with @AndreyFradkin.
Thanks also @erikbryn, @akorinek, and @professor_ajay for setting this all up :)
20 Oct 2025
(1/11)
How do AI agents change markets? We speculate on this topic in our new article for the NBER “Economics of Transformative AI” volume with @gilirusak, @BenSManning, @AndreyFradkin, and @johnjhorton:
“The Coasean Singularity? Demand, Supply, and Market Design with AI Agents”
2
5
29
8,682
Expected Parrot (YC F25) retweeted
20 Oct 2025
I have been waiting for a paper on AI agents and transaction costs and, well, agency problems
Agentic work, by its nature, drastically changes how these operate, with huge implications for how we organize markets and firms, which are largely shaped by agency & transaction costs
15
44
296
29,487
Expected Parrot (YC F25) retweeted
11 Oct 2025
You can just prompt things! And we want to make it easier for everyone to use these methods. As such, we're very excited for @ExpectedParrot to be part of the Fall YC Batch!
30
25
284
84,235
Expected Parrot (YC F25) retweeted
10 Oct 2025
Slide for a serious academic talk I'm giving on power calculations for experimental design via LLM powered simulation w/ @ExpectedParrot
5
4
72
7,139
Expected Parrot (YC F25) retweeted
8 Oct 2025
1
4
2,319
Expected Parrot (YC F25) retweeted
7 Oct 2025
Students in my class @LehighU have been using @ExpectedParrot to try finance tasks newly-enabled by LLMs. They've spun up "hedge funds" and directed "research analysts" and "portfolio managers" to learn about prop trading.
Last week, we did my favorite exercise yet.
27 Aug 2025
2
3
12
7,324
23 Sep 2025
[1/2] Forecasting questions can be tricky to ask — and even trickier to analyze — whether they’re posed to humans or AI agents. With EDSL, you can design forecasting questions in any format (ranges, rationales, or free text) and still get back a standardized dataset that’s easy to compare and analyze.
1
1
5
720
23 Sep 2025
[2/2] In our latest blog post and notebook we show how to:
🔹 Frame a forecasting question in multiple ways (structured vs free text)
🔹 Simulate answers with LLM-powered agent personas
🔹 Transform free-text responses (from AI or humans) into structured data automatically
👉 Check out the walkthrough here: blog.expectedparrot.com/p/de…
3
528
Expected Parrot (YC F25) retweeted
9 Sep 2025
This thread is amazing. A super clear demonstration of a great tool done respectfully with direct side-by-side comparisons. I found it directly persuasive to go look into @ExpectedParrot more.
9 Sep 2025
This is a cool & very interesting paper. It's great work. AND all the code to reproduce is available. The code is really well-done Python---perfectly clear & idiomatic. But I want to demonstrate how much time @ExpectedParrot can save researchers doing this kind of work 1/
4
17
8,351