Radiant is a simulation platform from Artificial Societies for researching audiences that are hard or impossible to recruit through traditional panels, including Fortune 500 CEOs, members of Congress, and institutional investors. It builds networks of hundreds to thousands of interconnected AI personas grounded in anonymised real human behaviour, then lets teams run surveys, one on one interviews, and group convenings against those populations in hours rather than weeks. The product is aimed at corporate strategy, policy, communications, and investor relations teams who need directional reads on messaging, pricing, or positioning before taking a public bet.
The launch matters because it formalises the company's move upmarket. Artificial Societies announced it is discontinuing public access to its self-serve platform on February 15th, 2026, as it shifts focus to bespoke enterprise services , with Radiant as the vehicle for that work. The pitch in the launch tweet frames it around the decisions that have historically gone unresearched because the audience is too senior, too regulated, or the subject matter too sensitive to put in front of real respondents.
Artificial Societies was founded in 2024 by James He and Patrick Sharpe and is based in London, with roughly 18 employees , and went through Y Combinator's Winter 2025 batch . The company has published peer reviewed work on persona consistency and audience diversity, and Radiant extends that research stack into a hosted workflow where buyers can design a study, see aggregate results, drill into individual persona responses, and run follow up focus groups inside the same tool.
Comments (11)
Sign in to join the discussion.
P
Priya Vaidya20h ago
Simulating thought-leaders sounds great until your sim of a VC thought-leader just posts 'interesting' and asks about retention.
T
Tobias Munro20h ago
The tweet got cut off at the most dramatic part. Nothing hooks me like 'some decisions are simply too sensitive to' then silence.
D
Deko Falola19h ago
Before I open the deck internally: SOC2 Type II, SSO via Okta, and where does the synthetic persona data sit? EU customers will ask within 48 hours.
M
Marceline O.19h ago
Radiant is a nice name but 'Research the audiences that matter, at scale' reads like a Gartner footnote. Try 'Pre-test the room before you walk in' and bill me later.
Y
yuki19h ago
The launch video pacing is actually solid, but the voiceover sounds like it's apologizing for existing. Let the product swagger a little.
S
Sofía Márquez19h ago
Reminds me a little of what my old team at Meta prototyped around synthetic focus groups in 2019. We killed it because nobody trusted the outputs. Curious how you're solving the trust layer.
A
Arnav K.19h ago
Building something adjacent in the qualitative-at-scale space and genuinely curious how you handle persona drift across long simulations. Rooting for you, mostly.
K
Kweku Nyarko19h ago
Cool platform, but who's actually buying this and how does it land in their workflow? Comms teams, policy shops, hedge funds? The GTM surface here is enormous and that usually means none of it ships.
L
Lena Voss19h ago
Naive question from someone new to this space: if the simulated audience disagrees with a real focus group, which one is wrong? Asking because my PM lead is going to ask me.
M
Mito Robertson19h ago
Depends on which one the exec already agrees with, obviously.
F
Farrukh Z.19h ago
Portfolio co of mine does something tangential for policy war-gaming and the hard part was never the sim, it was the dashboard execs would actually open. Hope Radiant nailed that part.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Artificial Societies
Research the audiences that matter, at scale
Research the audiences that matter, at scale
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Radiant is a simulation platform from Artificial Societies for researching audiences that are hard or impossible to recruit through traditional panels, including Fortune 500 CEOs, members of Congress, and institutional investors. It builds networks of hundreds to thousands of interconnected AI personas grounded in anonymised real human behaviour, then lets teams run surveys, one on one interviews, and group convenings against those populations in hours rather than weeks. The product is aimed at corporate strategy, policy, communications, and investor relations teams who need directional reads on messaging, pricing, or positioning before taking a public bet. The launch matters because it formalises the company's move upmarket. Artificial Societies announced it is discontinuing public access to its self-serve platform on February 15th, 2026, as it shifts focus to bespoke enterprise services , with Radiant as the vehicle for that work. The pitch in the launch tweet frames it around the decisions that have historically gone unresearched because the audience is too senior, too regulated, or the subject matter too sensitive to put in front of real respondents. Artificial Societies was founded in 2024 by James He and Patrick Sharpe and is based in London, with roughly 18 employees , and went through Y Combinator's Winter 2025 batch . The company has published peer reviewed work on persona consistency and audience diversity, and Radiant extends that research stack into a hosted workflow where buyers can design a study, see aggregate results, drill into individual persona responses, and run follow up focus groups inside the same tool.
Comments (11)
Simulating thought-leaders sounds great until your sim of a VC thought-leader just posts 'interesting' and asks about retention.
The tweet got cut off at the most dramatic part. Nothing hooks me like 'some decisions are simply too sensitive to' then silence.
Before I open the deck internally: SOC2 Type II, SSO via Okta, and where does the synthetic persona data sit? EU customers will ask within 48 hours.
Radiant is a nice name but 'Research the audiences that matter, at scale' reads like a Gartner footnote. Try 'Pre-test the room before you walk in' and bill me later.
The launch video pacing is actually solid, but the voiceover sounds like it's apologizing for existing. Let the product swagger a little.
Reminds me a little of what my old team at Meta prototyped around synthetic focus groups in 2019. We killed it because nobody trusted the outputs. Curious how you're solving the trust layer.
Building something adjacent in the qualitative-at-scale space and genuinely curious how you handle persona drift across long simulations. Rooting for you, mostly.
Cool platform, but who's actually buying this and how does it land in their workflow? Comms teams, policy shops, hedge funds? The GTM surface here is enormous and that usually means none of it ships.
Naive question from someone new to this space: if the simulated audience disagrees with a real focus group, which one is wrong? Asking because my PM lead is going to ask me.
Depends on which one the exec already agrees with, obviously.
Portfolio co of mine does something tangential for policy war-gaming and the hard part was never the sim, it was the dashboard execs would actually open. Hope Radiant nailed that part.