sheheryar (@_snaerr) header
sheheryar (@_snaerr) profile picture

systems guy. crackhead cto @n1chain.

nyc
Joined May 2023
Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)Photo by sheheryar (@_snaerr)
I do not send through “Jito”. I do not send through “Zeroslot”. I send to the CURRENT LEADER’S TPU, and if it drops, I kill myself. Albert Is Great
10
5
2
72
6,754
sheheryar retweeted
Asked @_snaerr where he & our intern went, should I be concerned??
5
2
18
810
sheheryar retweeted
N1's official job description for a rust engineer from our CTO. Ping me if you meet the criteria or know anyone. Referral bonus for a successful hire included 🫡
8
5
36
2,862
sheheryar retweeted
Replying to @chrislakin
reading non-fiction is a completely different activity. with fiction you read and let the story unfold like a movie. with non-fiction my reading speed is like 1/10th. i'm re-reading, thinking of examples, creating rebuttals, looking up alternative interpretations, and so on
1
1
5
349
sheheryar retweeted
90% of coding is waiting for a fever dream where the entire code is revealed to you
68
553
80
6,944
214,116
sheheryar retweeted
they need to run the goth mommy benchmark on o3. until we have those results, the ARC-AGI score is meaningless
2
2
1
30
1,324
sheheryar retweeted
Introducing Zeroth-01 Bot: the world's smallest open-source end-to-end humanoid robot starting at $350! Fully open-source, including hardware, SDK, sim environments & more. Zeroth-01 is the most hackable humanoid to-date! Join⬇️to shape the future of accessible robotics w/ us
We designed, 3D printed, and built a <$270, 13-DOF mini humanoid robot, wirelessly controlled and capable of walking and boxing at the @kscalelabs hackathon in 24 hours Open sourced: github.com/jingxiangmo/low_c…
25
201
34
1,174
145,950
sheheryar retweeted
Replying to @_snaerr @loqusion
radare2's command syntax 🤤
2
1
9
3,184
sheheryar retweeted
the eventually in "eventually consistent" means never
8
4
37
2,040
we remembered the twitter password!! in london, boston, waterloo, vancouver, montreal++ today — making retro game consoles, animations, material selection databases, stories, hacker fabs, and more
8
6
6
144
40,571
sheheryar retweeted
Builder aura in Montreal
I just launched a robotics hackerspace in Montreal! We're building a highly technical robotics community for researchers and engineers to work on robotics projects, research, and startups More info here: robotica.sh Open house this Saturday: lu.ma/sy2rd1do
1
2
1
54
5,317
sheheryar retweeted
human hands were not made to manage memory
2
1
1
16
1,218
3
3
50
2,435
sheheryar retweeted
Replying to @crypt0lake
we're getting there. improved tooling for reproducible fuzzing. newer languages secure by design. infrastructure becoming simpler and more powerful. plenty of startups in the area, just gotta look in the right places :)
1
1
272
sheheryar retweeted
who's going to rustconf come say hi
1
4
308
The decision to embrace VSR’s message passing for TigerBeetle was valuable and subtle: “Use the simplest building blocks you can, and then push control flow (and guarantees) up.” We learned this from networking, to follow the end-to-end principle. en.m.wikipedia.org/wiki/End-…
Why would Message Passing be superior to RPC for implementing consensus protocols?! Couldn’t we simply reason about remote procedure calls in terms of two messages being passed, a request and a response?! System Model We will think in terms of two layers, the application layer and the platform layer. The application layer consists of the parts of the system that are aware of consensus The platform layer consists of the parts of the system that are aware of sending and receiving messages Message Passing In case of message passing, the application layer is stateful, but the platform layer is stateless After sending a message, the platform layer has no expectations RPC In case of rpc, both the application layer and the platform layer are stateful After sending a request, the platform layer expects a response The problem The statefulness of the application layer contributes to finding consensus. However, the statefulness of the platform layer does not-at best, it is meaningless In consensus protocols, message routing, timeouts, or retries require application-level semantics that the platform-level is incapable of providing (E.g. increasing the value of a fencing token on retry) Conclusion Message passing is the lower level primitive on which remote procedure calls can be built. However, consensus is a great example where the lower level primitive is the better choice-the higher level primitive increases complexity instead of simplicity Message passing wins here teapawt.net/DominikTornow/status/1…
1
9
53
4,855
sheheryar retweeted
Replying to @kindgracekind
With age you stop doing that this much and suddenly you have simulated annealing
2
1
61
10,123
sheheryar retweeted
We designed, 3D printed, and built a <$270, 13-DOF mini humanoid robot, wirelessly controlled and capable of walking and boxing at the @kscalelabs hackathon in 24 hours Open sourced: github.com/jingxiangmo/low_c…
This quoted post is unavailable.
2
25
6
224
130,882
sheheryar retweeted
you know constructivist mathematics is gonna be good when an introduction to the field opens like this
3
6
1
37
3,130
Do you want formal verification for your actual system? Deterministic Simulation Testing is "dangerously" close teapawt.net/TigerBeetleDB/status/1…
This bug requires: - multiple view changes, along with - delayed disk writes, - concurrent requests, - successive crashes, and - perfect timings (!) to reproduce, plus assertions and online verification to detect. This is the magic of Deterministic Simulation Testing.
2
11
1,791