crypto quant | on-chain pvp | rated 9.999916/10 by hot market makers
Joined February 2021
- Tweets5.7K
- Following1.7K
- Followers12K
- Likes20.5K
Pinned Tweet
Trade six types of strategies
If you have momentum trades, add in a system for mean-reversion. If you have trend-following trades, add some carry strategies. +5 from one strategy, could be -4 from the other and it'll auto-reduce scale
h/t discussion w/ @SimpelAlpha
Thread👇
ALT Quant Crypto Trading Strategies
The native harness is much better than any custom harness you can make
You only need three things to manage a swarm and use the native harnesses:
1. Shared memory (e.g. your custom location or OptMem from @VictorTaelin )
2. Communication method (hijack Claude's SendMessage and Codex's queue)
3. Native subagents - e.g. spin up agents and initialize w/ tmux
Then for any fancy schedule/bot/manager/coordinator you can just use a long running session
Poor phrasing, this sounds like it’s the completely the engineers fault
What generally happens is attackers get into one service and start privilege escalation attacks until they reach the service they need
So it’s issues in the backend, just the entry point was the laptop
Binance's initial investigation suggests that the security incident may involve a compromised DevOps engineer’s laptop, which allowed an attacker to gain access to a service container and submit unauthorized withdrawal transactions. We have found no evidence of any wallet private key leak. We will share a detailed technical post-mortem once the investigation is complete and the facts are fully verified.
The estimated total impact sits at around $6.3B. We have worked with blockchain partners to freeze affected assets, with ~$3 billion frozen on-chain so far.
Binance will cover all losses from the SAFU fund and the company’s funds. User funds are safe. We will also publish an updated Proof of Reserves once the remediation is complete.
As an immediate precaution, Binance has:
- temporarily suspended deposits and withdrawals while security checks are completed;
- revoked the affected address authorisation;
- engaged law enforcement and blockchain intelligence partners to trace and recover funds.
Our teams are working around the clock to resolve this issue safely and thoroughly. We are working closely with our law enforcement partners to confirm the identity of these criminals and bring them to justice.
Please rely only on official Binance channels for updates. Please feel free to contact Binance Customer Support if you have any further questions.
binance.com/en/support/annou…
5.6B hack
Insurance fund 1.2B
Tether and Circle froze 2.2B
Means 2.2B uncovered
Currently - USDT withdrawals blacklisted and ETH withdrawals delayed/stalled
Binance was so incompetent handling this that @tether just froze Binance Hot wallet to prevent the hacker from sweeping more money Lmao
And good they did it as it prevented another $1M+ from being swept
This is why SOC2 Compliance is dumb
All the certificate says is: “I verified your system says what you claim it does”
So if your TOS claim is, “I will send all your information to North Korea,” you can get a SOC2 certificate
Checked out Lean, it's a functional programming language and you don't even need to use the theorem proof features if you didn't want to
Checked a basic computation benchmark, normalizing C++ to 1x, I got Lean 4x, Python 10x
From what I can see, fuzzing can get you 90% of the way there for practical programs within needing to use the full Lean proof framework
Oracle design is one of the hardest problems in defi
For this problem, Papertrade came up with a novel solution:
List it under Risk as "unresolved vector" and do absolutely nothing
wrong link but christ almighty yes it's a fking risk!
have to admit if it had more than 3 bucks in the LP i'd be tempted, i don't think it's even illegal
docs.papertrade.xyz/#/how/ri…
I looked through docs and the cost to trade makes the barrier pretty high, there's also some edge effects as LP bounces back and forth around 0
At some point everything is susceptible to a supply chain attack
If you go up the security levels, you're benefiting from making it harder for attacker to assume you're going to store crypto on it:
- Transparent case -> have to trust the board/chip
- BYOB -> have to trust the chip / tape-out
- Air gapped machine -> have to trust laptop sourcing and seed gen code
- Seed gen -> roll random dice into seed
oh my god this is worse than we thought
here's @Ledger $86M exploits explained
> cryptobilis was a reputable and authorized Ledger/Trezor/OneKey dealer for over 4 years
> they decided to sell the company to a Chinese guy (most likely patsy for DPRK) from northern China in MARCH 2026
> the "Chinese guy" modded the devices within the supply chain itself (physical), namely Nano X devices, to contain an implant module
> ledger validity/genuineness check CANNOT detect this because hw wasn't actually tampered with, just spied on
> quoting @MagicalTux:
" implant just monitors what's shown on the screen, will take note of the seed as it's shown to you, then transmits it to 3rd party"
> former owner chimes in on X (pic below) to say that "they were under a NDA" to not mention anything to anyone for 6 months
> they obviously didn't alert any of their partners that they were authorized to sell devices of... because NDA, duh!! its in the name!
> conveniently, funds start getting siphoned a few weeks before this magical NDA due date, and being laundered across all blockchains
Tired of seeing all the "wow inference is so profitable" takes
Everybody knows the trick to juice short term profit - just gut R&D
You don't get these inference margins w/o the model training costs to stay desirable
guy discovers how profitable inference is
Readers added context they thought people might want to know
The screenshot's Grok calc overstates inference profits by pricing all ~217k tok/s/GPU benchmark tokens at full output rates; AgentX traces show ~99.5% are cheap cache-hit inputs.
inferencex.semianalysis.com/run/deepseek-v… huggingface.co/datasets/semia… api-docs.deepseek.com/quick_start/pr… x.com/smallmodelsmel…
Dating sites should be government run
Moreso, they should make it illegal for others to run dating apps
Because of incentives: a gov’t wants you to get married and make more citizens
A private app wants you to stay single, unhappy, and paying forever
Bend is interesting because it puts the invariants up front
Currently w/ LLM, the way to use it is to have very strong CI-CD and validation checks (end to end to avoid regressions)
So this flips that around - put the checks up front so you can't write it wrong in the first place
You just have to look at the machine they are using, its diff per lane:
- Tier 1 machines can differentiate water and bomb, so liquids not problem
- Tier 2 machines have 3D view so can keep laptop in
- Tier 3 machines - all out
After 2026, shoes always on, belts always off
Micro-shorts similar to this is very popular in the brainrot part of the internet
How it makes money:
- Each episode is a few minutes long and it ends on a massive cliffhanger (free to watch)
- The creators then sell season-subscriptions where users can have all the episodes at once
If @goodalexander bakes in cliffhangers into every single one of these, he will become an unstoppable media empire
My team asks how I name our servers and i tell them I use a random name generator
But the secret is my daughter keeps a list of names of her rubber ducks and i just go down the list
This post has one sentence reminded me of the good old days:
> going to bed falling asleep thinking about the same problem every night
Now it’s been replaced with triggering “just a few more” overnight run before bed
I recently crossed 1 year working at OpenAI.
I reflected on doing good work, company culture, and the 5 biggest surprises from an action-packed year:
aalokpatwa.substack.com/p/on…
The work that Takopi put into quantifying Kalshi wash trades is very similar to the work
that you would do as a quant to harvest a bad-MM
multi-market-movement monitor, similarity of quotes
See pdf below - github.com/OctopusTakopi/kal…
