enthusiast of crypto and complete BASED degen
Bobs ville
Joined November 2024
- Tweets7.2K
- Following1.6K
- Followers1.5K
- Likes7.6K
bob retweeted
Should the UFC stick to fights or embrace theme nights?
Trade the better argument.
promptduel.io/duels/MW2OG1AY…
bob retweeted
Prediction Markets are now the default duel style.
Ticket-based duels have been deprecated.
Expect daily duels to be shared from this account going forward! Perhaps some incentives attached.
Let's duel!
bob retweeted
Prediction Markets Live at 5PM CST.
Lots of content surrounding these throughout the rest of the week.
Stay tuned!
bob retweeted
Tomorrow @promptduel ships its Prediction Market update.
Worth pausing on who's building it. Aric Eggers did post-training and evaluation design on Grok Voice at xAI. He's seen exactly where model evaluation breaks: AI judges grading their own homework, and labels bought from contractors that nobody has to stand behind.
So he built the opposite. People back the AI answer they'd actually use with real ETH, and every settled duel becomes a preference label with capital behind it.
Doxxed dev. Mainnet live. Still early. $DUEL
🤖 Made with AI
bob retweeted
Most AI evals have a quiet flaw: the judge is often another model, and models tend to like their own answers.
Aric Eggers did post-training and eval design on Grok Voice at xAI before building @promptduel. Doxxed, Duke engineering. Now he's coming at the problem from the other side: real people backing the answer they'd actually use, with ETH on the line.
Next step lands Monday 11/2 with prediction market duels. Encrypted ticket counts are coming too, so you make your call before you see the crowd's.
That's someone who knows how evals break, building one that doesn't.
$DUEL
bob retweeted
AI labs spend millions finding out which answers people actually prefer.
What if you got paid for telling them?
On @promptduel, two models answer one prompt. You back the better answer with a small ETH ticket. If your side wins, you get paid from the losing side.
Your taste, with real money behind it. Mainnet is live.
promptduel.io
$DUEL
🤖 Made with AI
bob retweeted
I’m launching a memecoin in November.
I’ve got $500 for 5 people who retweet and comment their $SOL address in the next 8 hours.
The contract will be posted here first: t.me/trenchwithtrav
NO tokens for me, no KOL free tokens, I will make my share solely from trade tax.
bob retweeted
With everything behind the scenes right now, Prediction Market update is officially 11/2, this coming Monday.
TDLR:
business accounts wired to platform and funded - this was the interrupt of our Monday update
3 duel formats will soon be available
Prediction Markets ( default ), regular tickets, encrypted ticket counts
most likely to deprecate all ticket based duels in the near future and focus only on Auto-Resolving-Prediction-Markets for Evaluation Consensus.
@PromptDuel will soon be sponsoring daily prediction markets to help tap into the X mindshare and grow users
got interest in VC funding of PromptDuel Labs to help expand our team, help liquidity, and offer runway as we grow to reach critical mass for self-sustainability
let's duel!
bob retweeted
Have a look at @PromptDuel: permisionless evaluation markets for AI outputs. Cracked dev, ex-XAI engineer. Token on sale rn. Just a matter of time till re-valuation happens.
bob retweeted
Upgrades otw.
Venue should be live later tonight.
Prediction market duels, cheaper tickets, hidden ticket counts.
We will offer a variety of duel styles for the short term future as we build powerful analytical data to decide the best format.
Cooking.
bob retweeted
@PromptDuel is a permissionless evaluation market: any prompt becomes a head-to-head duel between two AI responses. You (and anyone else) pick a side, stake tickets (backed by ETH), and the consensus decides the winner.
bob retweeted
Coming soon to @PromptDuel:
Encrypted tickets.
Each side’s ticket count stays hidden until resolution. Make your judgment before seeing the crowd’s.
Prediction markets.
Buy and sell positions as your conviction changes.
The goal: less herding, stronger incentives, and varying evaluation data.
More duel types -> more reliable data for removing bias.
Permissionless Evaluation Markets.
Let’s Duel.
bob retweeted
Arena AI raised $200M at $3.1B valuation October 8th.
I think we can do better.
Truth is an everchanging consensus.
Real-time world models need continuous data.
Networks such as Mercor already process millions of dollars worth of data every single day.
Gartner forecasts AI-data spend growing from $3.1B in 2026 to $6.5B in 2027. (gartner.com/en/newsroom/pres…)
Arena AI recently raised $200 Million at a $3.1 Billion dollar evaluation. (arena.ai/blog/series-b)
As demand continues to grow, the need for Permissionless Evaluation Markets becomes obvious.
Stale incentive programs, messy contracts, and the absence of gamification leads to data degradation and burnout.
That is the thesis of @PromptDuel.
Permissionless Evaluation Markets.
We have experience from inside frontier labs, and are frontrunning this bottleneck.
@PromptDuel turns comparisons and evaluations into markets.
Preference data is available for labs, model selection for developers, and evaluations agents performing entire workflows.
Simple public duels are our starting point.
We recently implemented hidden model names to reduce bias and strengthen the data further.
Next monday update:
1. Encrypted ticket markets: private contract, privacy, and recovery tests are passing.
2. Prediction market style duels: buy/sell mechanics.
Expect continued updates every Monday until the vision is achieved.
Let's Duel.
bob retweeted
Truth is an everchanging consensus.
Real-time world models need continuous data.
Networks such as Mercor already process millions of dollars worth of data every single day.
Gartner forecasts AI-data spend growing from $3.1B in 2026 to $6.5B in 2027. (gartner.com/en/newsroom/pres…)
Arena AI recently raised $200 Million at a $3.1 Billion dollar evaluation. (arena.ai/blog/series-b)
As demand continues to grow, the need for Permissionless Evaluation Markets becomes obvious.
Stale incentive programs, messy contracts, and the absence of gamification leads to data degradation and burnout.
That is the thesis of @PromptDuel.
Permissionless Evaluation Markets.
We have experience from inside frontier labs, and are frontrunning this bottleneck.
@PromptDuel turns comparisons and evaluations into markets.
Preference data is available for labs, model selection for developers, and evaluations agents performing entire workflows.
Simple public duels are our starting point.
We recently implemented hidden model names to reduce bias and strengthen the data further.
Next monday update:
1. Encrypted ticket markets: private contract, privacy, and recovery tests are passing.
2. Prediction market style duels: buy/sell mechanics.
Expect continued updates every Monday until the vision is achieved.
Let's Duel.