Just Learning, Growing, and Showing up Everyday!

Sparta
Pinned Tweet
Been using the same pfp & banner for a long time, so I thought it was time for a little refresh. > Updated both > Made it a bit cleaner What do you think, is it good?
154
4
336
17,929
just woke up and found out Bitget got hacked for $351.6M 💀 here's what happened so far: - Bitget detected unauthorized transfers at 18:31 UTC - around $351.6M in assets were affected - the breach was limited to parts of its hot + warm wallet infra - Bitget says its cold wallets remained secure - withdrawals are currently paused while the security review is ongoing - deposits is still operational - Bitget says user balances remain accurate and user funds are protected - Bitget CEO says Lazarus Group is suspected to be behind the $351.6M hack - the same group was linked to the $1.46B Bybit hack in 2025 - the attacker spoofed transaction data and triggered the authorization process to move funds - private key compromise has been ruled out - they said the loss is contained and no further unauthorized transfers are possible $351.6M is insane. if you already have funds there, don't panic. withdrawals are under review, so wait for the official update from Bitget. that's why : if you don't need to trade it, keep your funds in a cold wallet.
87
2
113
1,537
the NFT you grinded hard for gets you rugged the one you missed cooks hard and the one you hold is almost dead this game is rigged!
91
2
112
844
finally back after a smol hibernation and seems like i missed a lot of drama
85
115
852
happy weekend gng looks like many got approved by @zaddrnet mine is still in review, dunno if it's gonna get approved or rejected tho anyways, it's weekend go afk for a bit and touch some grass!
112
134
1,044
dunno if i'm late or what to apply for @zaddrnet but finally got around to applying. - 2800 PFPs on Zcash - built around shielded ownership through the Orchard pool - public face, private owner the WL is already getting crazy 17k+ applications for 2,800 spots definitely gonna be a tough one let's see if i make the cut or get cooked this time
Nice to not meet you. 2,800 faces on @Zcash . Every face is public. Every owner isn't. Whitelist is open. Apply early. zaddr.net/apply
53
1
81
780
so today is the last day of the @axisrobotics Kaito campaign. before it ends, let's recap a few recent updates: Axis × Feagine Robotics: > Axis just announced a commercial partnership with Feagine > a company building soft embodied intelligence infrastructure. > simulation task design + trajectory generation > training data for Feagine's Fi0 cross-embodiment foundation model then we have the $AXIS Community Sale: > $1M total raise > $0.10 per $AXIS > 10M $AXIS / 1% of total supply > $100M FDV > commitment window: Sep 21–28 > 10% unlock at TGE. and with that, I'm marking this as my 6th and final post for Epoch 1. so far, 5.3K+ creators have participated in the campaign. once this epoch ends, Kaito will publish the Lb. let's see it goes. Btw have you submitted your top 6 posts yet?
from no one's posting about Axis to everyone's posting about Axis. @axisrobotics kaito campaign is ending in 3 days and we've already got 5.2K+ creator participants. and honestly, the growth around Axis has been pretty hard to ignore lately. the Hub is now showing: > 5.4M total robot trajectories > 200k+ total users > 5,034,260 transactions that's a lot of activity for a platform where contributors are actually generating robot training data. then there's the Axis × Binance Keyless Wallet Alliance: > 3.4k participants > 270k trajectories > 1.5M Axis Points reward pool the campaign has its own task pool, with up to 20K tasks released daily. and then we have the Axis contribution roles. a recent announcement confirmed: > total active role holders across all three tiers are now under 750 members > this includes X-Axis, Y-Axis and Z-Axis roles so these roles are becoming a lot more limited as the ecosystem grows. --- and Axis just crossed 5M robot trajectories on Base, generated by 200K+ contributors worldwide. now only a few days left for Kaito. it's gonna be pretty impressive to see where this ends up. and I guess everyone's participating at this point but, if you're still fading, just hop in: s.kaito.ai/VncNymW
77
3
109
995
day one of Arc was full of controversies - many ppl got rugged by NFTs and memecoins - the whole "mint on a private site" drama - multiple NFT mints went live yesterday and honestly, almost 90% of them were just waste - some retards are dragging Arc devs based on nationality - and the most hyped NFT, Arclings, turned out to be a clown i was eligible with 2 wallets and couldn't mint a single NFT the entire time because the website was dead. also lost a few moni on memecoins lol but minted @akadotfun $50 -> $230 they did a great job with the launch. didn't start off great for Arc on day one, but somehow ended with a W. how was your day one on Arc?
71
5
93
1,699
finally, @arc mainnet is going live today new chain, fresh liquidity, new users & everyone trying to figure out where the early alpha is. bridge is already open ahead of the launch: - bridge.usdc.com/ - relay.link/bridge - jumper.xyz/?toChain=5042 there are multiple launchpads, NFTs and memecoins coming on Arc and obviously only a few will perform. @Arguspad and @TollyLabs are two launchpads everyone's keeping an eye on. once the gates fully open, expect a lot of FOMO some existing tokens could get a serious volume rotation if they catch attention but don't be that guy aping every random ticker just because it says Arc verify the ca, check the liquidity, look at the holders and DYOR. early doesn't automatically mean Ws. Gl!
69
1
98
846
from no one's posting about Axis to everyone's posting about Axis. @axisrobotics kaito campaign is ending in 3 days and we've already got 5.2K+ creator participants. and honestly, the growth around Axis has been pretty hard to ignore lately. the Hub is now showing: > 5.4M total robot trajectories > 200k+ total users > 5,034,260 transactions that's a lot of activity for a platform where contributors are actually generating robot training data. then there's the Axis × Binance Keyless Wallet Alliance: > 3.4k participants > 270k trajectories > 1.5M Axis Points reward pool the campaign has its own task pool, with up to 20K tasks released daily. and then we have the Axis contribution roles. a recent announcement confirmed: > total active role holders across all three tiers are now under 750 members > this includes X-Axis, Y-Axis and Z-Axis roles so these roles are becoming a lot more limited as the ecosystem grows. --- and Axis just crossed 5M robot trajectories on Base, generated by 200K+ contributors worldwide. now only a few days left for Kaito. it's gonna be pretty impressive to see where this ends up. and I guess everyone's participating at this point but, if you're still fading, just hop in: s.kaito.ai/VncNymW
93
2
124
2,267
KRATOS retweeted
Wanted: The Sherwood Six 🤠 30 GTD spots on the table. 5 per head. To claim a spots: → Follow us → Like + RT → QT this post, tell them what we're building, tag the wanted you're hunting → Submit form below 48 hours then the sheriff rides ⭐
940
2,865
4,175
162,413
Wanted: @Tma_420 Found the Exit Leader I'm hunting. @exitlabsdotfun is building a founder game around 4,444 Offices on Rh Chain. You start with an unfunded Office, activate it with $EXIT, then work your way through: - Garage - Shared Space - Seed Office - Unicorn HQ - Exit Tower Higher stages come with higher earning rates, and eventually you decide when it's time to EXIT. That's what caught my attention. The Office isn't just a collectible, it actually has a progression system behind it. @Tma_420, your bounty is mine!
Wanted: The Sherwood Six 🤠 30 GTD spots on the table. 5 per head. To claim a spots: → Follow us → Like + RT → QT this post, tell them what we're building, tag the wanted you're hunting → Submit form below 48 hours then the sheriff rides ⭐
54
72
780
happy sunday gng minted Royal Mechanica: $8 -> $50 ath shill me some more W NFTs need something new to ape into anyways, got any plans for today?
what if a successful robot trajectory isn’t actually useful training data? sounds weird, right? but that’s what @axisrobotics tested. they collected 660 human correction attempts for a robot task. but only 161 eventually made it into training. the filtering looked like this: - 660 corrections collected - 496 successfully replayed - 416 passed the redundancy check - 414 reached short-snippet testing - 161 passed the final recovery test so 253 of those 414 didn't pass the final recovery test. the reason is pretty interesting. a human completing a trajectory doesn’t prove that their correction actually fixed what the robot was failing at. Axis extracted the short correction, then handed control back to the original policy. if the robot could finish on its own after that correction, the snippet was useful. if not, it was discarded. each useful snippet was only around 0.8 sec long, or 4 control commands at 5 Hz. and the result? the original policy managed 66/160 successes. with the 161 verified snippets, that went to: 78/160 80/160 84/160 across three different training seeds. the takeaway isn't just “collect more data.” it’s figuring out which data is actually worth learning from. 161 snippets made the cut. now i'm making the cut for the weekend. go touch some grass.
93
125
1,506
what if a successful robot trajectory isn’t actually useful training data? sounds weird, right? but that’s what @axisrobotics tested. they collected 660 human correction attempts for a robot task. but only 161 eventually made it into training. the filtering looked like this: - 660 corrections collected - 496 successfully replayed - 416 passed the redundancy check - 414 reached short-snippet testing - 161 passed the final recovery test so 253 of those 414 didn't pass the final recovery test. the reason is pretty interesting. a human completing a trajectory doesn’t prove that their correction actually fixed what the robot was failing at. Axis extracted the short correction, then handed control back to the original policy. if the robot could finish on its own after that correction, the snippet was useful. if not, it was discarded. each useful snippet was only around 0.8 sec long, or 4 control commands at 5 Hz. and the result? the original policy managed 66/160 successes. with the 161 verified snippets, that went to: 78/160 80/160 84/160 across three different training seeds. the takeaway isn't just “collect more data.” it’s figuring out which data is actually worth learning from. 161 snippets made the cut. now i'm making the cut for the weekend. go touch some grass.
90
6
135
3,302
my $6 and a dream are gone now @world_xyz said they'll relaunch it but we've already seen what happened with the MoonPay relaunch congrats to everyone who got their stimmy anyways, wishing you all a good day we move.
99
1
129
1,602
The Goated trio when it comes to farm community
42
63
633
if you missed a few updates from @axisrobotics this week, here's the weekly wrap-up. there’s quite a bit to catch up on. [ what’s new? ] New Dataset + Real2Sim2Real: > bringing real objects and scenes into simulation > generating high-fidelity data in sim > checking on hardware if the data actually helps > focusing on cross-object and cross-scene skills like pick-and-place, packing and sorting Active Curation: > moving beyond human, VLM and script-based checks > using model performance to decide which trajectories are actually useful > putting the training budget toward data that can still teach the model something Online Post-Training: > repeated rollout failures become the next collection targets > operators take over, correct the failure and hand control back > new corrections are mixed with older data so existing skills don't get erased Verification + Anti-Cheating: > full action replay > manual cross-review > bots and cheating operators are promptly banned Dual-Arm Support + New Tasks: > dual-arm support with ARX5 > multiple task descriptions in the same scene > tighter length, width and height requirements > new put-in and take-out basket tasks Axis already has large-scale data generation and collection in place. now the focus is on using model performance to figure out what data should come next. and 9 days left on the campaign, so there's still time to make a few good posts and see where it takes.
69
5
94
846
new week again So what we clicking this week?
112
135
987
apparently @world_xyz is cooking something dunno what, dunno if the rumour is even real but the whole TL is suddenly desperate for that Fb we've seen some crazy drops on Sol before could be early access, NFTs, a drop or something else dunno what they’re cooking But I'm still sidelined tho.
47
66
971