Carsten Munk @stskeeps

wannabe manfred macx || co-founder @directionallyai || exocortex maximalist Warsaw, Poland Joined May 2009

Tweets

33K
Followers

4K
Following

7K
Likes

11K

Talip @otaliptus

9 hours ago

Find me around if you want free sats

fun 12 hours for @lxdev_ after fixing some last minute bugs, we're onboarding our first users at @BTCPrague. @otaliptus or @stutxo are on the ground if you wanna test unrelated photo of zerosats' first payment to a bark wallet ⚡️

1 2 28 1K 2

1 2 9 678 0

View Details

Daria @DariaStrategy

a day ago

Stop trying to meet people. Start giving them a reason to meet you. Build a product, a podcast, a community, an event, a newsletter. Something that puts you at the center of a conversation you care about. Then keep showing up long enough for the relationships to become real. It's slow and unsexy, but it's the only thing that works. I've launched a YT channel with tech philosophy for people who hate tech content. In the first video, I share my unconventional playbook for building relationships in global tech. What helps you to build connections?

2 2 11 260 3

View Details

Boris Cherny @bcherny

a day ago

Just landed nested subagent support in Claude Code Starting to experiment more with agents kicking off agents as a way to better manage context. Capped at depth=5 to start, going out in today’s release. Lmk what you think!

493 293 6K 454K 1K

View Details

nairolf @0xNairolf

2 days ago

real question: out of all apps that launched perps after hyperliquid, how many have actually been somewhat successful? zero?

42 17 85 9K 0

View Details

Carsten Munk @stskeeps

2 days ago

there's a difference between: knowing and applying at the right moment

0 0 0 38 0

View Details

Carsten Munk @stskeeps

2 days ago

might be in the meme mining business

0 0 0 48 0

View Details

Wizz Air @wizzair

3 days ago

RIP Airplane Mode. @Starlink onboard internet is coming to WIZZ IN 2027.

307 276 9K 2.6M 565

View Details

Carsten Munk @stskeeps

3 days ago

@theo mosh?

0 0 0 105 0

View Details

Nyakio @nyakiomaina11

3 days ago

so, what is zkTLS and how does it work? notable work in projects like, @zkPass @reclaimprotocol @tlsnotary

4 10 46 929 5

View Details

calle @callebtc

4 days ago

Thoughts on unruggable Cashu mints in trusted execution environments (TEEs). Sharing current research and looking for feedback. Idea: Generate Bitcoin reserve keys and Cashu signing keys inside a TEE so operators never access them. Goals: - Minimize rug risk (ideally to ~0) - Push operators toward non-custodial status # Risk 1: Malicious mint updates Users can verify the software running inside a TEE, but operators could later deploy a malicious fork. Potential mitigations: - Infrastructure provider enforces which software may be deployed. - Separate KMS with policies from the TEE allowing only audited/signed software to access keys. - Immutable mint: disable updates entirely. Upgrades require deploying a new mint and users migrating funds, similar to smart contracts. # Risk 2: Rollback attacks ("time travel") An operator could restore an old database snapshot, allowing previously spent ecash to be spent again and inflating supply. Possible mitigations: - Append-only storage enforced by the TEE/provider so spent notes can never be erased. - Seal the database inside the enclave and prevent external access. Combined with update protection, this appears to eliminate rollback attacks. # Risk 3: Mint shutdown The operator can always turn the mint off, making reserves inaccessible. Potential solution: @lukechilds proposed an elegant emergency exit mechanism. Reserve UTXOs contain a secondary spend path that activates after a period of mint inactivity. A third party (e.g. a multisig) holding a copy of the mint database could continue honoring withdrawals. During normal operation the mint periodically rolls reserves forward, extending the timeout. If it stops, the emergency path activates. A more ambitious direction is committing the spend book as a Merkle/nullifier tree. With the right scripts, users may be able to withdraw unilaterally by proving their ecash has not been spent. This could largely solve risks 1-3 simultaneously. # Conclusion We haven't cracked this nut completely yet, but the progress over the last few weeks has been remarkable. The mere existence of plausible solutions is incredibly exciting. Imagine if we could pull this off. David Chaum invented ecash in the 1980s, and researchers spent decades trying to make it trustless. Then Satoshi invented Bitcoin, and the world largely moved to the blockchain paradigm. It would be an enormous achievement if we could finally solve the trust problem of Chaumian ecash by rebuilding it on Bitcoin and its more expressive L2s, combining the best of both worlds: a scarce, immutable base layer with trust-minimized ecash on top, offering strong privacy, instant payments, near-zero fees, offline transactions, and amazing UX.

14 23 150 11K 37

View Details

Daria @DariaStrategy

4 days ago

99% of networking events feel fake. I've launched a YT channel with tech philosophy for people who hate tech content. In the first video, I share my unconventional playbook for building relationships in global tech. Share your tips for making friends in new environments.

2 2 10 277 2

View Details

Taelin @VictorTaelin

5 days ago

This has been a massive success! So I left 32x GPT 5.5 agents (enough to fill the 5h limit) on 32 separate machines. Each one received HVM5's unoptimized file, and a prompt demanding for a 10x speedup. After 4 rounds, all agents reached ~2x speedups, with 1.3x to 2x increases in file sizes. Many cool ideas surfaced. Some are obvious and most agents rediscovered to them, like computed gotos. But others are actually surprising. For example, turns out a bounded LIFO freelist is MUCH faster than the usual algorithm. Many agents didn't bother trying. Neither did I! Now I'm in the process of mining insights from the 32 runs so I can merge them into a "super HVM5". Not sure what is the best way to do this, and my 200x credits are about to expire, but I still have a few hours, so I have one shot at getting it right. Sounds fun! (Also thanks for the free credits, I appreciate it a lot...)

Taelin @VictorTaelin

6 days ago

Since my 200x Codex credits end tomorrow, I've redirected the Mac Mini cluster to optimize HVM5. Depending on the time it ends, I should have 6h to 30h of dozens of GPT 5.5 agents working on it. I've wrote the most careful anti-reward-hack prompt ever. Let's see how it goes!

25 4 508 138K 84

36 37 1K 176K 408

View Details

Miles Cranmer @MilesCranmer

5 days ago

This is an insane paper and I love it arxiv.org/abs/2605.31514

159 1K 11K 615K 6K

View Details

Kamal Gupta @kamalgupta09

5 days ago

1 2 57 6K 3

View Details

Carsten Munk @stskeeps

5 days ago

given all the wild VC stories, what does an agentic post-VC world look like? working capital on observable product fit? investor agents entering TEE with agent(s) from the company?