frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Mavera – Predict audience response with GANs, not LLM sentiment

https://docs.mavera.io/introduction
4•jaxline506•2d ago
Mavera is an audience intelligence API. Give it a message, product prototype, or creative asset and it returns a predicted distribution of emotional and behavioral responses across your target stakeholder population. This is the best way to test your assumptions before you spend or push anything live.

To show this in practice, we ran all 101 Super Bowl LX ads through Mavera on game night: https://superbowl.mavera.io. We simulated how audiences would emotionally and behaviorally respond by platform and segment. We returned a distribution rather than a single score as part of a full analysis of each ad in under 4 hours.

The model is a GAN adapted for language, emotion, and cognition. A generator produces synthetic audience responses and a discriminator validates them against human benchmarks. Scoring follows a feel-think-act framework: emotional activation, cognitive framing, behavioral prediction. We validated scoring against the Harvard/Illinois OASIS benchmark. MAE on emotional response is 0.02-0.15 versus 1.0-2.5+ for GPT and Claude. Every response includes a confidence score and a hallucination risk score. You can also build-in spread of opinion, response stability, and impact of news/market context scores to your outputs.

The API is OpenAI-compatible. Change the base URL to app.mavera.io/api/v1, add a persona_id, and you are running against 50+ pre-built personas or you can customize your own. Sub-100ms latency at P99. Free API key and docs at https://docs.mavera.io/introduction.

Comments

jaxline506•2d ago
Most "AI ad testing" is GPT sentiment scoring with a wrapper. We built something architecturally different and the Super Bowl felt like the right moment to show it publicly. The core issue is that LLMs model language. They predict what a person might say about something, which is not the same as modeling how a person will respond. That requires thought-to-behavior mappings, not next-token prediction. We call our scoring framework FEEL-THINK-ACT: emotional activation, cognitive framing, and behavioral prediction in that order. To check we were not just building a different flavor of hallucination, we benchmarked against the OASIS dataset (Harvard/Illinois). Our MAE on emotional response is 0.02-0.15 vs 1.0-2.5+ for GPT/Claude, emotional accuracy 98% vs ~78% for base models, consistency across runs 96% vs 72% for competitors. Every output also carries a confidence score and a hallucination risk score because we did not want to hide uncertainty behind a clean number. On the simulation side we do not model a single synthetic person. We run the target population 10,000 times with built-in variation and return a distribution. The tails and the variance are the insight. For the Super Bowl we scored all 101 units pre-game through post-game covering memorability, brand clarity, audio-off performance, and cultural resonance by platform and audience segment in under 4 hours. Live context came in via our 615 Environmental Intelligence system which folds news cycles and cultural signals in at run time, so the scores reflect the world as it was when the ads aired. The API is OpenAI-compatible. If you are already building on OpenAI it is a base URL swap:

from openai import OpenAI

client = OpenAI( api_key="YOUR_MAVERA_KEY", base_url="https://app.mavera.io/api/v1" )

response = client.chat.completions.create( model="mavera", messages=[{"role": "user", "content": "Score this ad copy for emotional resonance."}], extra_body={"persona_id": "YOUR_PERSONA_ID"} )

print(response.choices[0].message.content)

Free tier, no enterprise contract, no demo call. Full methodology and scores at superbowl.mavera.io, API docs and free key at docs.mavera.io. Happy to dig into the OASIS benchmarking, simulation architecture, or 615 in the comments.

hayksaakian•1h ago
How did the predicted response compare to actual responses for super bowl ads?

I noticed the benchmark mentioned but its too jargon-filled to follow for me.

troelsSteegin•19m ago
What were the personas [0] trained on?

[0] https://docs.mavera.io/features/personas#persona-types

Astral to Join OpenAI

https://astral.sh/blog/openai
859•ibraheemdev•5h ago•536 comments

An update on Steam / GOG changes for OpenTTD

https://www.openttd.org/news/2026/03/19/steam-changes-update
73•jandeboevrie•1h ago•41 comments

Show HN: Three new Kitten TTS models – smallest less than 25MB

https://github.com/KittenML/KittenTTS
120•rohan_joshi•2h ago•32 comments

How to Not Pay Your Taxes

https://taylor.town/succession-000
39•surprisetalk•1h ago•19 comments

UK's Ofcom has today fined 4chan £450k for not having age checks in place

https://www.ofcom.org.uk/online-safety/illegal-and-harmful-content/4chan-fined-450000-for-not-pro...
31•longislandguido•1h ago•23 comments

OpenBSD: PF queues break the 4 Gbps barrier

https://undeadly.org/cgi?action=article;sid=20260319125859
129•defrost•4h ago•40 comments

World Happiness Report 2026

https://www.worldhappiness.report/ed/2026/
52•ChrisArchitect•2h ago•38 comments

Juggalo Makeup Blocks Facial Recognition Technology (2019)

https://consequence.net/2019/07/juggalo-makeup-facial-recognition/
180•speckx•5h ago•88 comments

Launch HN: Voltair (YC W26) – Drone and charging network for power utilities

17•wweissbluth•1h ago•0 comments

Noq: n0's new QUIC implementation in Rust

https://www.iroh.computer/blog/noq-announcement
11•od0•16m ago•0 comments

Prompt Injecting Contributing.md

https://glama.ai/blog/2026-03-19-open-source-has-a-bot-problem
55•statements•2h ago•20 comments

Return of the Obra Dinn: spherical mapped dithering for a 1bpp first-person game

https://forums.tigsource.com/index.php?topic=40832.msg1363742#msg1363742
15•PaulHoule•2d ago•0 comments

The Shape of Inequalities

https://www.andreinc.net/2026/03/16/the-shape-of-inequalities/
61•nomemory•3h ago•7 comments

macOS 26 breaks custom DNS settings including .internal

https://gist.github.com/adamamyl/81b78eced40feae50eae7c4f3bec1f5a
201•adamamyl•3h ago•102 comments

Android: Balancing Openness and Choice with Safety

https://android-developers.googleblog.com/2026/03/android-developer-verification.html
22•0xedb•1h ago•2 comments

Scaling Karpathy's Autoresearch: What Happens When the Agent Gets a GPU Cluster

https://blog.skypilot.co/scaling-autoresearch/
29•hopechong•1h ago•7 comments

Launch HN: Canary (YC W26) – AI QA that understands your code

14•Visweshyc•2h ago•11 comments

Afroman found not liable in defamation case

https://nypost.com/2026/03/18/us-news/afroman-found-not-liable-in-bizarre-ohio-defamation-case/
918•antonymoose•8h ago•511 comments

4Chan mocks £520k fine for UK online safety breaches

https://www.bbc.com/news/articles/c624330lg1ko
58•mosura•3h ago•43 comments

I turned Markdown into a protocol for generative UI

https://fabian-kuebler.com/posts/markdown-agentic-ui/
18•FabianCarbonara•4h ago•9 comments

Love of corporate bullshit is correlated with bad judgment

https://pluralistic.net/2026/03/19/jargon-watch/
41•hn_acker•1h ago•9 comments

Consensus Board Game

https://matklad.github.io/2026/03/19/consensus-board-game.html
58•surprisetalk•4h ago•9 comments

Hyper-optimized reverse geocoding API

https://github.com/traccar/traccar-geocoder
38•tananaev•3h ago•9 comments

Connecticut and the 1 Kilometer Effect

https://alearningaday.blog/2026/03/19/connecticut-and-the-1-kilometer-effect/
3•speckx•41m ago•0 comments

What if Python was natively distributable?

https://medium.com/@bzurak/what-if-python-was-natively-distributable-3bfae485a408
32•bzurak•3d ago•15 comments

Conway's Game of Life, in real life

https://lcamtuf.substack.com/p/conways-game-of-life-in-real-life
288•surprisetalk•14h ago•77 comments

Ramtrack.eu – RAM Price Intelligence

https://ramtrack.eu
55•nu11r0ut3•5h ago•18 comments

Show HN: Dumped Wix for an AI Edge agent so I never have to hire junior staff

6•axotopia•2h ago•3 comments

Monuses and Heaps

https://doisinkidney.com/posts/2026-03-03-monus-heaps.html
9•aebtebeten•18h ago•1 comments

Eniac, the First General-Purpose Digital Computer, Turns 80

https://spectrum.ieee.org/eniac-80-ieee-milestone
95•baruchel•12h ago•39 comments