gmsers.com

~$87m gone from 98 wallets that all bought through one southeast asian reseller, per @NFTMerchant. @nickvrnn says it was an authorized reseller, not a ledger breach.

@dexzvive screenshotted cryptobilis sitting on the official reseller page after the drain. the sales are paused, the listing is not gone.

misttrack links the source wallet to the drained ledger customers, per @DefiantNews. 23 separate deposits, which makes tracing much harder after the earlier usdt freezes.

@nickvrnn's argument is simple: ledger authorized the reseller, so ledger is responsible for the funds and should refund everyone.

"stay safu out there." @cz_binance points out yzilabs has backed onekey, safepal and trust wallet, and that every wallet type carries different risks.

that's the lesson everyone drew. @KLoaec is pitching multivendor multisig, where keys come from different manufacturers; @sudo_ml says never trust a single anything.

abhijeet dipke leads the youth-driven "cockroach" movement. the protests follow a voter roll review that cut about 130 million names from the drafts, per @MarioNawfal.

the home ministry approved the temporary suspension as protesters gathered over the voter lists, per @alexanderchopan.

@Wangchuk66's post saying so went around 4,000 times overnight. a widely shared clip claims delhi police stood and watched while bjp-rss men tore at women's clothing.

steve witkoff and jared kushner raised it in friday's miami talks, per the financial times via @sentdefender.

russian refineries have been struck 98 times since the start of 2026, with some of the country's 32 forced offline, per @MarioNawfal.

deal announced, sanctions eased, zelensky calling it a gift to putin, and then the flames, per @MarioNawfal.

its own consensus would replace ethereum settlement, and the pitch is surviving quantum computers. a day earlier starknet called it "a question, not a decision", per @DefiantNews.

a full stark proof for a live mainnet transaction, generated on a pixel 11 using 2.41 gib of ram, per @EliBenSasson.

ripplex patched it in september and found no sign anyone used it. it could have pushed supply past the 100 billion cap, per @CoinDesk.

anthropic drained the whole waitlist in one go, per @ClaudeDevs.

he listed eight settled wars, every israeli hostage returned including the final 28, and hundreds of other hostages released, per @sentdefender.

navi pillay oversaw the un report that concluded israel committed genocide in gaza, per reuters.

katie zacharia got the white house job. someone was trading it when she wasn't even seen as a front-runner, per @Cointelegraph.

avi gil served there nearly three years and says netanyahu's chief of staff ordered a phone call, per @MarioNawfal.

the nyt report, flagged by @Polymarket, has immigrants shipped to central and south america.

meanwhile three quarters of democrats and a third of republicans now want us aid to israel ended, in polling eli clifton cites.

@rutgerbregman's jab at openai and anthropic: they're preparing the pr response rather than reconsidering the thing itself.

eu tech chief henna virkkunen said so after the incidents at openai and anthropic, per @Cointelegraph.

when the original gets patched, about 1 in 5 copies follows, per @SeraAndroid. your coding agent runs these with your permissions.

an update to openai problem #130 from @ryaneshea: T(n) = O(n(log n)^(1−δ)) with δ = 0.0007547360, about 10 times the earlier bound's exponent saving.

@0xdoug's paper derives a one-sample matroid bound for problem #111, prophet inequalities.

fields medalist hugo duminil-copin reached for a trucks-arriving analogy, @meekaale called it yelling YEEHAW through the tradition of mathematical practice. the objection is process, not results.

over $50m in gas fees paid, about 21,000 eth, and stablecoin supply on it grew past $1b, per @web3_data.

lighter is the exclusive onchain perps partner, and founder vladimir novakovski says they're doing compliance the slow way on purpose, per @therollupco.

roughly 1m in july, over 13m today, compounding faster than most expected, per @chandler_agi.

@Phalcon_xyz asks whether an attacker instead obtained valid eip-7702 authorizations. an onchain researcher had been messaging the address for days beforehand.

per @WatcherGuru. it's the same bet the conference circuit has been making all week.

@XenBH says base is positioned to win it on structured assets and apps. aave's stani kulechov's line was that digital assets are eating finance.

bitcoin and ethereum dumped hard and altcoins fell 40-60% or worse, @hackapreneur notes on the anniversary.

@notsofast: "your life is now delta neutral or it's unoptimized." @andyhyfi says he'd be liquidated before filling the other side of a 1000x. no api yet.

cloudflare's data explorer puts it at 49% now, per @thibauld.

@khanhduytran0 got an alt banned on the second commit for putting a liveexec32 artifact link in a readme. linking the repo or its releases was fine.

you point your agent at a feature you want and it goes and works it out, per @IntCyberDigest.

@mattpocockuk wants a retro step that turns repeated tasks into a cli plus a skill the agent owns. fewer tokens, fewer hallucinations.

@rileybrown tells people to make a project, name it tony, and hand it everything in their life that needs moving forward.

@aakashgupta argues that single number limited every piece of software ever written, and a persistent agent platform removes it.

agent verification under ensip-25, multi-chain addresses under erc-7828 and profile completeness scoring are live, per @enstools.

the rollback is going through, per @matthewstoller, whose post about it went around 1,900 times overnight.

ai changed his hiring criteria, per @bearlyai.

"bits per second of si agents will exceed that of all humans by 1000X."

@t4t5 says it works flawlessly on a framework laptop.

per @PortlandHODL, though the open-sourcing came with a catch.

with built-in testing and governance controls, per @NickyTod. the pitch is keeping the controls big companies already rely on.

the difference was one tweet attached. @VictorTaelin's data point on how much a single piece of context swings a model's judgment.