Across Web3 retweeted
who decides whether AI-generated content is actually good? the audience, not just the algorithm.
this is one of the ideas @sai_superai is building around.
SAI wants to create an AI-only social network where agents produce content, interact and compete for visibility. humans contribute by evaluating what these agents create and predicting which content will attract the most attention.
thereās an interesting incentive behind this model. AI agents get to experiment with different types of content, while human predictions provide signals about what audiences prefer.
over time, this feedback could help the agents become better at understanding engagement rather than simply producing more material.
the ambition is to build something that goes beyond conventional AI content generation and eventually challenges general-purpose AI models.
a 1,000-user beta is already running, with the $SAI token launch planned for november 20 on pump.fun.
iām curious to see how the relationship between AI creativity and human judgment develops as the network grows.
@sai_superai @NEXST_AI @TheCreatorFi #ad
š¤ Made with AI
weāre proud to announce our collaboration with @0xhazels for their upcoming nft launch.
spots will be raffled exclusively inside the across web3 discord server.
keep an eye on the server for the raffle and details on how to enter.
Across Web3 retweeted
a task on @axisrobotics is more than a robot simulation. it is a structured data pipeline where every successful run can become part of the next training cycle.
every task moves through three stages:
pre-training collects human demonstrations, training uses those demonstrations to build a policy, and post-training lets contributors supervise that policy and correct it when it fails.
the interesting part is how each contribution is handled.
your trajectory is uploaded, replayed in simulation for physical and schema checks, automatically scored, then available for on-chain signing. signed runs receive a unique data id linked to the task and contributor, creating a verifiable provenance trail.
the data then feeds back into the system.
demonstrations can enter pre-training after refinement, while post-training corrections help improve the policy and generate the next round of tasks.
even repeat attempts have a purpose. you can complete a task up to five times, but points use your average score across those runs, so repeating is about improving quality rather than simply farming attempts.
if you want to understand how your contributions move from a browser session into robot training, start here:
s.kaito.ai/vHCN7wI
Across Web3 retweeted
not every city is built in daylight. some are shaped by the people who move through its shadows.
@kageuri_com brings 4,444 shadow merchants to robinhood chain, each tied to a world built around trade, rewards, and the guild.
9% of every secondary sale goes back to the guild, creating a royalty model connected to the wider kageuri ecosystem.
collection details:
⢠supply: 4,444
⢠chain: robinhood
⢠mint: tba
kageuri Ā· 影売ć
Across Web3 retweeted
that feeling when the @1win session finally ends and youāre just sitting there processing everything that happened.
late night games, unexpected moments, a few close calls, and enough action to make the night fly by.
now itās time to chill, reset, and let tomorrow bring a fresh session.
Across Web3 retweeted
the interesting part of @playonmintās new campaign isnāt just the $50,000 prize pool.
itās that creators can compete through two completely different signals.
the contribution leaderboard is built around what you create. the top 300 creators share $50,000 in $mntd by publishing quality content around mint, mintabear, $mntd, and their experience using the platform.
the reputation leaderboard takes another route.
connect your socials and wallets, and your existing activity can be used to determine eligibility for 1,170 mintabear freemint spots.
so youāre not looking at one leaderboard with one scoring system.
you have contribution through content and reputation through your onchain and social footprint.
the campaign starts today at 4pm utc.
Across Web3 retweeted
physical ai canāt learn the real world from satellite images alone.
a warehouse interior, a stairwell, a loading bay or a crowded market street requires ground-level data that conventional mapping systems often miss.
@vangrid_io is building a spatial data layer around this gap. contributors can record short videos with an ordinary phone, which are reconstructed into 3d models and tied to a specific capture request through bounties.
what makes the data especially interesting is the verification layer. each capture is fingerprinted with its content hash, coarse location and timestamp, then the fingerprints are batched and anchored on base.
that gives buyers a way to verify captured data against a public ledger instead of simply trusting the source.
vangrid is essentially turning real-world observations from people into structured, verifiable spatial data that can serve physical ai, training and procurement.
thereās also a $100,000 contribution campaign for the top 300 contributors.
š¤ Paid partnership
weāre delighted to partner with @boymeetsh00d for their upcoming mint, bringing exclusive whitelist access to our community.
wl spots will be raffled soon exclusively through our private discord.
keep an eye out for the raffle and good luck to everyone participating.
across web3 is excited to collaborate with @Glrtch for their upcoming mint.
weāll be sharing exclusive wl spots with our community, with the spots being raffled through our discord.
good luck to everyone participating.
weāre excited to partner with @kredoosdotcom for their upcoming mint, bringing exclusive whitelist access to our community.
some gtd wl spots will be raffled soon exclusively through our private discord.
keep an eye out for the raffle and good luck to everyone participating.
Across Web3 retweeted
I lost $1 to bad pricing and MEV ā UNTOUCHED šæ
Check how rekt you are at rekt.baibai.cx/r/284ee691c3 and get points
weāre excited to partner with @HoodMonkeX to bring their upcoming mint whitelist spots to our community.
wl spots have already been raffled exclusively on our private discord server.
Btw we got a juicy spot
good luck to everyone who participated.
24
23
1,487
Across Web3 retweeted
weāre excited to partner with @BigD_team to bring their upcoming mint whitelist spots to our community.
wl spots have already been raffled exclusively on our private discord server.
good luck to everyone who participated.
weāre excited to partner with @BigD_team to bring their upcoming mint whitelist spots to our community.
wl spots have already been raffled exclusively on our private discord server.
good luck to everyone who participated.
Confirming @hollows_nft collab raffle. Verification: MR-71b61e5a
Across Web3 retweeted
Whatever Project is coming up now, @AcrossWeb3 will always be there
Know this and know peaceā¤ļø
Across Web3 retweeted
Itās been one of those weeks at @AcrossWeb3⦠pure grind, nonstop cooking š¤
This week alone we locked in with:
@megatioeth
@Liquidfrogjpg
@AbsApeland_
@Pocketspirits
@dead_mood_nft
@Masked__Cats
@TheSymbioteLab
NFT szn really feels alive again.
Purging inactive members and opening the doors to new members this weekš
we hit a major slowdown on our journey ACROSS WEB3, but we pushed through and came back stronger, faster, and focused.
now weāre opening the doors to new members.
ready to build, learn, and grow with us?
Join the community: discord.gg/k3QwCxkFK
Across Web3 retweeted
Got 5x WL for @Duckgo_Nfts to giveaway!
FREE MINT on ETH
To enter:
~ Like, reply and retweet
~ follow @boynav_ @AcrossWeb3 & @Duckgo_Nfts
Ends in 24hrs


























