coffee
8.8K posts
Companies with the most lifers (> 10 yrs at company) and their median pay
English
The only reason to publish a photo(?) this grainy in 2026 is to obfuscate things lol
Unless there's a higher resolution one that I havnt seen yet, he's 100% incapacitated or dead
dart@poordart
Survival rate after out-of-hospital cardiac arrest by age
English
@coffeedev they have... or do you want trucks to go straight from any location to the final destination?
English
@RhysSullivan Never once did I think the sam altman company would be the less evil of the ai labs but anthropic seemingly cannot stop doing everything they can to screw their users
English
@P3b7_ @Ledger @DonjonLedger Who needs a $250k laser when you've got a $5 pipe wrench and the leaked name and address of every ledger customer!
English
💥One Check, One Laser, Every Card: The Tangem Immutability Trap.
The @DonjonLedger just published research worth stating plainly.
With a single laser pulse the Ledger Donjon team faults one conditional check in the firmware of a Tangem card and resets the password to a value of its choosing. No existing password, backup card, or recovery feature needed. Once reset, the attacker is able to sign anything and can potentially drain the user’s wallet.
To be precise about the threat model: this attack requires physical possession of the card, invasive chip opening, lab-grade fault-injection equipment, our own setup costs roughly $250,000, and genuine technical expertise. That puts it well beyond the reach of an opportunistic thief, but comfortably within the capabilities of a serious lab. Both things can be true at the same time.
Why one pulse is enough: the recovery path depends on a single yes/no check, “is this card in recovery state?”, and a pulse is simply a precisely timed electrical disturbance designed to make the chip misread that decision at the critical moment. Because there is no redundant check and no penalty for repeated SetPin attempts, one successful disturbance is enough. The chip’s countermeasures are formidable, but they cannot protect the one bit the firmware chose to trust.
The uncomfortable part: it cannot be patched. Tangem cards have no firmware update mechanism. The vulnerability was disclosed on February 10th, 2026. There is no fix coming, because there is no channel to deliver one.
Tangem presents immutable firmware as a security feature. Call it what it is: a trade-off. "We cannot change the firmware" is a strong story right up until the firmware is wrong. Then the same property that protected you guarantees you can never be protected again. This research did not create that reality. It made it visible.
What a user can do, since there is no patch: this attack requires physical possession of the card and invasive lab work, so it cannot be done covertly. The practical risk is a lost or stolen card in the hands of a capable attacker. If the card stays in your possession, there is no reason to assume compromise. If you have doubt, or if your threat model requires a higher level of assurance, treat the funds as compromised and move them to a new secure set up.
This was published in line with Ledger Donjon’s responsible disclosure process. When a vulnerability cannot be patched, the next responsibility is to inform users clearly and widely, so they can make their own informed decisions, especially here where the vulnerability can not be exploited remotely.
The bigger lesson is not about one product. Security is never static, and systems should be designed with human error and future failure in mind. You should not have to blindly trust that yesterday’s assumptions still hold. You should be able to verify, adapt, and recover when they do not. Design for the day you are wrong, because eventually, you will be.
Full write-up from Baptistin Boilot, Ledger Donjon.
Stay safe. Stay honest about your trust assumptions.
donjon.ledger.com/blog/bypassing…
English
@nftgoy @ProofOfEly @Abstract_Eco so, if someone sells when they are up they are demonized for dumping on people's heads
if they sell when they're down/round tripped they are laughed at and called a paper hands
nice little lose/lose you manage to create for everyone that wants to participate in something
English
@duncanstives @kevworld86 They will all be losers in the end. The only winner will be the lawyers
English
@kevworld86 The biggest loser will depend on the results of engineering analysis. If it was a defect in construction then the contractor is gonna have a bad day, if it's a defect in design then the engineering firm will be fucked.
English
As a guy who deals with capital construction projects, the lawsuits that are going to come out of this will be phenomenal. The developer is going to sue both the contractors and architects/engineers. The contractor is going to counter sue everybody. wsj.com/real-estate/ma…
English
Apparently I'm allowed to talk about GPT-5.6 now?
It's a damn good model. Not quite as "smart" as Fable, but it is incredibly capable. Fixed all the problems I had with GPT-5.5.
It is incredibly determined. Will run for a day without even using a /goal. It understands subagents incredibly well and is great at orchestrating. It's super pleasant in use cases like OpenClaw and Hermes Agent. It knows iOS dev incredibly well.
It has rough edges too, but FAR fewer than 5.5 did.
For many things, gpt-5.6-sol will become my obvious default.
I will share a lot more soon 🫡
English
@MarcosHernanz Paseo is phenomenal - been using it as my daily driver for a while now
English
@betty_nft @backseats_eth Pretty sure it is a prerequisite to be here in the first place
English
@adamscochran I mean, he's been brain dead for years already so it's not exactly news
English
Imagine you’re Mitch McConnell’s staffers.
Chronically plugged in to every pulse of the hill.
It’s been 5 days since rumors of your boss’ health started to spiral
Major papers are now asking where he is.
Some pundits declare he is brain dead.
And you decide… not to address it?
You decide to put out no new statement.
You decide Mitch, who you claim is hard at work and spoke last week to Thune, won’t tweet or write a press release.
And when asked by the press for a statement, you give the same canned response from 3 days prior.
Because that’s how you handle this totally normal and benign situation that’s totally not a problem…
English
Hearing a big announcement is on the way before the end of the day. If it is what I think it is, it’s absolutely massive.
English
@cryptovcdylan @DrNickA @lzminsky the price of diem doesn't affect the cost or utility value at all though - even if diem dropped to $0.1 on the market, the total liability of venice is capped at diem supply dollars worth of inference credits.
English
English
On Venice and VVV
This is worth digging in to because there is both a legitimate argument that there is a pervasive misalignment between equity and token holders systemic across the space, but also a clear demonstration of a utility token in the wild.
There are two ways to get access to inference on the Venice platform, you can just buy inference from them with a credit card, or crypto and just spend those credits.
Or, you can stake VVV and mint DIEM an asset that gives you a budget of inference every day. This is one of the clearest examples of a utility token we've seen. It's a good token model.
This means that there is both a real buy side demand on the token for something other than just basic speculation, or barely existing governance rights. To get that utility you lock it up, lowering the effective float of the token, which means that demand for the utility has a greater impact on the market price. A genuine reflexive flywheel. I am convince that this token popped earlier in the year during the agent wave because of this. That's why it's up in a bear market.
The whole point of venice is that you can get anonymised access to AI inference. This is the major point. It means that I can set up a wallet (a pseudonymous identity), capitalise it with crypto and run AI inference without doxxing to the overlords. That's a credible reason for both crypto, and a token model which has a pay off profile if I use the inference every day. End-to-end crypto access to AI, absolutely crucial for personal agents unless you literally want to download your life into Claude's mind.
So, you have two aspects of the business an openrouter style on demand inference / compute company and a token economy that offers a product for routine ongoing inference. It's perfectly reasonable to assume that the company running that has a business is worth owning a stake in, ~and~ that there is a token economy worth participating in, either directly for its inference utility, or because you think other people will do it and you want to bet on the adoption of that product.
Now, you can moan that you didn't have access to the equity and there is no reason why they can't tokenise it and have a dual token model in the market. But there is a viable reason for both of those assets to exist.
Utility tokens are a real thing. They are not equity. It's ironic that after the hundreds of raises where teams didn't even bother building a token model and just raised on tokens as proxy equity that this is the project everyone gets on their high horse about this issue.
Tokens can be way more than simply equity, they can have a diverse utility case. They can be a currency in a network economy, that both accrues value based on the adoption of the product through direct access to revenue via buybacks and the like, but also additional perks, benefits and access of the product, or products in a decentralised ecosystem.
Now, would I be happier if only tokens existed. Absolutely. But that means doing a completely alegal punk af raise into a non-wrapped DAO and issuing tokens from there. Based. But we don't do that do we.
English
@Rhynotic @heppe_eth have you tried getting audits before deploying contracts instead
English
we’ve brainstormed a fix. the plan is to redeploy with the bugfix and (more improvements) after an audit this week
ill also open-source the new contracts before deployment and let any whitehats try to break it for a bounty
stay tuned
Adam@Rhynotic
Looks like there was a way to frontrun the chainlink callback, allowing an attacker to shift which NFT was selected. As a result, they managed to purchase someone's CryptoPunk unfairly (the largest listing in the pool). I'm in contact with the owner of Punk 5450 and TokenWorks will cover the ETH (~$66k). We've set the protocol into withdraw-only mode as of block 25452023 (and snapshotted $FWA holders here as well), preventing purchases and allowing depositors to withdraw their assets. I'll need a few days to decide what to do next, and to see if there is a path forward where the project can continue/relaunch since I personally think everyone is having fun with it (when it works). Sadly this is the downside of launching new things semi-quickly. Thank you for the support and for trying our ideas 🫶
English
@MikeSilagadze Curious why this needs a whole instance - isn't the hub and spoke model of V4 designed to be able to support this without a fully separate deployment?
English
fable gets removed from the sub on 7th, gpt 5.6 releases on 7th or 9th, fable miraculously comes back
Thariq@trq212
I've heard a lot of questions about Fable's availability on subscription plans. While it will come off subscriptions after July 7th, we aim to restore Fable as a standard part of our subscriptions as soon as capacity allows, as we mentioned in our original blog post.
English