Shahar Papini Profile
Shahar Papini

@PapiniShahar

2,282
Followers
300
Following
13
Media
1,139
Statuses

Works at StarkWare

Joined May 2018
Don't wanna be here? Send us removal request.
@PapiniShahar
Shahar Papini
1 month
@DrBriefsScratch I don't understand why you don't just send your CV and apply for jobs. There are a bunch of hiring platforms and companies accepting CVs. With your resume you get a lot of interviews, and probably nail a few.
25
0
498
@PapiniShahar
Shahar Papini
2 years
Who wants to help me build the first ever zkEVM, on top of Cairo? Seeiously though, I had some POC in a local branch a while ago, it seems feasible to me.
30
15
198
@PapiniShahar
Shahar Papini
5 months
Prediction: Circle STARK based proof systems will be the fastest in the world. I expect we'll hear a lot more about this in the coming months.
@StarkWareLtd
StarkWare 🐺🐱
5 months
Math-heads, you’re in for a treat. Introducing Circle STARK 🔵 Over the past three months, we’ve worked in collaboration with the Polygon Labs team on a new blazingly fast STARK protocol. Today, we are proud to announce a ground-breaking mathematical advancement: ✨ A
Tweet media one
78
135
756
10
23
151
@PapiniShahar
Shahar Papini
7 months
@ShaykhSulaiman How can anyone believe a word Hamas says? We all saw videos of Hamas propaganda where the "bodies" literally came to life. Claims about the hospital bombings that turned out to be false.
15
1
110
@PapiniShahar
Shahar Papini
1 year
Me: New zk paper! Nice! Paper: We introduce the Glabglab system, which uses Gazorbazorg commitments with a sprinkle of Trpeztel circuits (Shtick[17])
7
7
97
@PapiniShahar
Shahar Papini
4 months
People ask me why we have our own Circle STARK implementation (Stwo), separate of plonky3. Stwo's main and only goal is to accelerate Starknet. Through fast cairo and recursion proofs. This concreteness allows us to advance fast and write optimized code.
2
12
86
@PapiniShahar
Shahar Papini
3 months
Would you like Starknet to support llvm (i.e. write your code in native language, say, rust), even if the cost overhead was x3?
19
9
69
@PapiniShahar
Shahar Papini
5 months
8
5
66
@PapiniShahar
Shahar Papini
2 years
Finally, a compiler from Idris (Haskell-like pure functional language) to Cairo! Also works with starknet:)
2
17
63
@PapiniShahar
Shahar Papini
10 months
@tc_modestine @ChayaRaichik10 Sounds more like the Mars Attacks scene where they are welcoming the hostile aliens for the *second* time, attributing their massacre to "misunderstanding"
2
0
57
@PapiniShahar
Shahar Papini
11 months
@EliBenSasson comparing FRI to EC based commitment schemes.
Tweet media one
3
10
53
@PapiniShahar
Shahar Papini
10 months
@davidrkadler 260 civilians mirdered just in the musical festival, not even including the families murdered at their homes, or the kidnapped.
10
1
47
@PapiniShahar
Shahar Papini
6 months
Amazing features! And all I had to do was move to another project! Kudos to our wonderful compilers team
@Starknet
Starknet 🐺🐱
6 months
Cairo v2.5.0 was just released! While loops, storing strings, moar match flexibility, pub & pub(crate), useful warnings, and as usual, various small improvements. Community forum post: Github:
46
131
752
5
2
41
@PapiniShahar
Shahar Papini
1 year
Mind blowing! This is essentially an opensource prover for starknet! Well done!
@andrewmilson
Andrew Milson
1 year
Stoked to announce I just submitted my own proof to StarkWare's StarkNet/StarkEx verifier on Ethereum. 🧵👇
31
73
408
1
3
40
@PapiniShahar
Shahar Papini
6 months
@skdh The problem is with human language. It is ill defined ambivalent and exploitable. We should start talking in math, with well defined terms and implication rules.
17
1
43
@PapiniShahar
Shahar Papini
1 month
@crackticker The definition of bad code: code that someone else wrote
0
1
42
@PapiniShahar
Shahar Papini
3 months
Me and the guys on the prover team made this Circle Stark prover called Stwo, and it has nice performance.
@Ingo_zk
Ingonyama
3 months
Did you miss @StarkWareLtd 's talk from ZK Accelerate?😱 Watch it here ~ live from ZKA Athens @PapiniShahar presented: STWO Performance 🔥 YouTube: #zkaccelerate #zk #StarkNet
4
15
57
3
4
37
@PapiniShahar
Shahar Papini
10 months
ML framework in rust! I think rust is a lot easier to use than python, personally (and I have professionally used python for years)
1
0
30
@PapiniShahar
Shahar Papini
2 years
Skyro v0.2 just released! With #Starknet support too! "This release features StarkNet support, enabling developers to write StarkNet contracts in Idris2." ().
4
6
26
@PapiniShahar
Shahar Papini
8 months
@MaskNewCEO @visegrad24 You obviously have never been to Israel
2
0
25
@PapiniShahar
Shahar Papini
4 months
@IroncladDev Reported for dehumanization
0
0
25
@PapiniShahar
Shahar Papini
11 months
There will be some very interesting talks in San Francisco!
@Starknet
Starknet 🐺🐱
11 months
What a line-up is waiting for us at the Starknet Summit: Top researchers, Starknet projects, and the latest updates from the most innovative initiatives in the ecosystem - Starknet, Ethereum, and other L2s. 🔥🍸🦭✨ Check out the agenda in the following post.
40
254
607
0
1
21
@PapiniShahar
Shahar Papini
2 years
Some thoughts I had regarding code safety. Two competing ways to statically enforce code invariants are: 1. Static analysis. 2. Invariant by construction. It might be tempting to use existing constructions (such as existing programming languages), and throw some analysis on it.
1
2
21
@PapiniShahar
Shahar Papini
2 years
Thoughts about blocks and states. IMO, there is a slight issue in ethereum API. Both getBlockByNumber() and getBalance() get a block number, but the second actually refers to the state *after* that block, not to the block itself.
2
4
23
@PapiniShahar
Shahar Papini
5 months
OG
0
1
23
@PapiniShahar
Shahar Papini
3 months
@Anthony_Bonato To the point. All papers should be one pagers.
0
0
21
@PapiniShahar
Shahar Papini
14 days
@dimahledba @VitalikButerin I think it's commendable how quickly @VitalikButerin grasped the theory and practice, and implemented a circle stark prover in just a few days. Brilliant fellow.
2
0
23
@PapiniShahar
Shahar Papini
1 year
@m_ou_se The size of the target folder is a big issue. It's hundreds of gigs in our project...
0
0
20
@PapiniShahar
Shahar Papini
4 years
Tweet media one
0
3
18
@PapiniShahar
Shahar Papini
3 years
@0xmons @nickryg @iamDCinvestor @nour_noire @ABTestingAlpha ImmutableX will be using cairo, as the rest of our products do (deversifi, dydx, etc)
1
2
18
@PapiniShahar
Shahar Papini
9 months
@YosephHaddad מי שחשב שהיא נשארה בחיים, חי בסרט.
14
0
17
@PapiniShahar
Shahar Papini
3 months
Having a junior review your code might even be better than a senior. The goal is to make simple code understood by all.
@StephenPHassall
Stephen Paul Hassall
3 months
As a senior would you want a junior to review your code?
406
5
383
2
2
17
@PapiniShahar
Shahar Papini
1 year
2
1
16
@PapiniShahar
Shahar Papini
10 months
@yanisvaroufakis It's only a siege to get our innocent kidnapped citizens back. We will protect our people. But Israelis do not want palestinians to die. When they are brought back, this siege will be lifted.
6
1
16
@PapiniShahar
Shahar Papini
12 days
@untitled01ipynb Unpopular take: TF 1.0 was awesome, and did what it intended to do. A compiled computation graph framework not only for ML. TF 2.0 was trying to pivot and be something it's not. "more of the same", and killed TF.
2
0
19
@PapiniShahar
Shahar Papini
5 months
This guy almost spilled the beans
@CPerezz19
CPerezz.eth ⟠ 🦀
5 months
LatticeFold, Beyond Circuits and now by @UHaboeck The explosion of papers of this end of February is being insane! Will try to schedule L&S sessions for the following months with the authors if possible :)
5
7
48
1
1
14
@PapiniShahar
Shahar Papini
11 months
I know kung fu
2
0
9
@PapiniShahar
Shahar Papini
11 months
@fede_intern @jdiegooliva @Starknet Lambdaclass has done a wonderful work in StarkNet. Major contributor to the great state we are in right now. We wouldn't be where we are without you. Thanks for all your work!
1
0
13
@PapiniShahar
Shahar Papini
22 days
@fede_intern You know how much backlash I got for this? We tried to tell people..
@PapiniShahar
Shahar Papini
11 months
@EliBenSasson comparing FRI to EC based commitment schemes.
Tweet media one
3
10
53
1
1
14
@PapiniShahar
Shahar Papini
2 years
Come hear me talk about the next evolution of Cairo
@StarknetCC
StarknetCC
2 years
Don't have your tickets for StarkNetCC, but still want to go to the event? @StarkWareLtd , our main sponsor, is allocating 50 tickets to the community so that everyone could enjoy the event! 👇
10
8
27
0
3
13
@PapiniShahar
Shahar Papini
5 months
Yes please!
@fede_intern
Fede’s intern
5 months
wow Shahar, one of the creators of Cairo use to play the game mod created by the company we bought. Let’s find an excuse to add Cairo to it 🫡.
0
0
7
0
0
11
@PapiniShahar
Shahar Papini
4 months
@weikengchen Well, we'll see on Thursday at my talk on zkAccelerate:)
0
3
12
@PapiniShahar
Shahar Papini
1 year
@p_e @GuthL @milancermak Louis, you are literally a member of Starkware and I'm sure I told you this a million times when discussing typed starknet and sui style txs. You are just bsing!!
1
0
12
@PapiniShahar
Shahar Papini
3 years
Some fixed point operations in cairo
@guiltygyoza
guiltygyoza
3 years
Let's do physics on #StarkNet
6
11
71
0
3
12
@PapiniShahar
Shahar Papini
8 months
@peter_szilagyi So, you basically say, we should "trust" the devs. I think he raises a valid (known) point. No one is pointing fingers. It's the principal.
1
0
12
@PapiniShahar
Shahar Papini
4 years
@hkalodner @PhABCD @StarkWareLtd @reddit @ethereum We only publish each vault's total change onchain. Even though there were 300k txs, there are only 100k vaults, each is changed thrice. We need to save the public key (32B), token id(2bit with a lookup table for the 3 tokens) and balance (8B). ~14B per tx
1
3
10
@PapiniShahar
Shahar Papini
1 year
@liamzebedee @p_e It's the plan, but not soon
0
0
9
@PapiniShahar
Shahar Papini
1 year
@fede_intern Are you trying to take over the world???
1
0
10
@PapiniShahar
Shahar Papini
4 years
@Chloethedev @flovanlieshout @StarkWareLtd @reddit @ethereum The solution presented in the post has no data availability committee. All the data is onchain. i.e. It's rollup mode, not validium.
0
2
10
@PapiniShahar
Shahar Papini
1 year
@milancermak Yikes... Please don't... Use logic, not blind beliefs...
0
0
10
@PapiniShahar
Shahar Papini
1 year
@p_e @raphael_dkhn 1. Prefer to use Span in stead of " @Array ". 2. Use span.pop_front() for iteration - a lot more efficient.
1
0
9
@PapiniShahar
Shahar Papini
11 months
In slightly more words: To show inclusion of one column in another, logUp verifies sum_i 1/(z-a_i) = sum_j m_j/(z-b_j), at a random z. We can express this sum as an arithmetic circuit using projective coordinates (to avoid division), and use the GKR protocol to prove it.
1
1
9
@PapiniShahar
Shahar Papini
2 years
Bravo!
@dimahledba
A₿del ∞/21M 🐺 - 🐱
2 years
And we dit it! 100% EVM opcodes compatibility on @KakarotZkEvm ! In 2 months and 20 days... No funding, no company, only a community of passionate builders ❤️ I am so thrilled. Still a long way to go before running into production but this is amazing 🔥
Tweet media one
34
120
511
2
0
9
@PapiniShahar
Shahar Papini
3 years
@LanrayIge @ukolodny @epolynya I suspected you had this misunderstanding. dydx and immutable are not in starknet. They are starkex instances. If/when they move to starknet in the future, they will be composable.
1
0
9
@PapiniShahar
Shahar Papini
5 months
@itamarl @tarrenceva @CairoLang @class_lambda World day of Social Justice, of course! Wouldn't miss it fml.
0
0
9
@PapiniShahar
Shahar Papini
23 days
@nenosipapuce @StarkWareLtd @EliBenSasson 512 bit -> 256 bit. Standard hash security with 128 security bits. Same payload.
1
1
8
@PapiniShahar
Shahar Papini
6 months
@nesua_plus יש תופעה שאדם (בייחוד נשים) נמצא במצב "משותק" ולא מסוגל להגיד "לא". הבעייה היא שהמון אנשים (בייחוד גברים) לא מודעים לתופעה הזו. למיטב ידיעתם, האישה יכלה פשוט להגיד לא. צריך ללמד את הציבור על התופעה, ולהסביר שעקב כך תמיד חייב "כן" מפורש.
2
0
8
@PapiniShahar
Shahar Papini
2 years
@0xlowki @0xFranzKafka Starkex derivatives is a Starkware product. The cairo code was written by a Starkware team of talented individuals (Ilya, Shahak, Dan, and more..), and myself.
3
0
7
@PapiniShahar
Shahar Papini
2 years
@mayazi We had 2 choices: 1. evm emulator. Fully compatible, but potentially slower. 2. Evm transpiler. Not as compatible. Faster. Path to migrate to pure cairo. We chose the second.
1
0
8
@PapiniShahar
Shahar Papini
1 year
@sylvechv @eniwhere_ @Starknet Actually, it was @LiorGoldberg2 . I'm just taking all the credit
1
0
8
@PapiniShahar
Shahar Papini
1 year
@gaetbout Auto completion for methods. Interoperability testing.
1
1
8
@PapiniShahar
Shahar Papini
1 year
@feltroidPrime @dimahledba Theoretically you can write your hint functions in Cairo1 itself, and have a single libfunc that calls a function as a hint. I always had this vision of writing hints in Cairo itself. Maybe it'll even be supported in starknet if it's safe enough.
2
2
8
@PapiniShahar
Shahar Papini
1 year
I'm new to React, but I feel like useRef is an antipattern. Everything should be a pure function of the state. useRef is a state outside the state. Also, I don't like useContext. It makes passing values implicit. I believe that better design of param objects can replace it.
3
0
7
@PapiniShahar
Shahar Papini
11 months
Thus, no need to commit on the 1/(z-a_i) column - which has big and extension field values. We also show to use GKR and sumchecks with univariate schemes (like FRI)
1
1
8
@PapiniShahar
Shahar Papini
2 years
@VitalikButerin @EliBenSasson @Justindr @TimBeiko @dankrad That is exactly the issue. Referencing states by blocks is doomed to fail. There are more states than blocks. States are the nodes of the tree, where blocks are edges (state transitions) Example: Take the first block id. Does it represent a block? If so, there was a state before
2
0
6
@PapiniShahar
Shahar Papini
2 years
@BowlOfChilli Yes, EVM interpreter on starknet. Which is roughly equivalent to zkevm, right?
3
1
6
@PapiniShahar
Shahar Papini
2 years
@0xAurelou @milancermak felt will still exist
2
1
7
@PapiniShahar
Shahar Papini
2 years
There are a lot of cases when developers encounter invariants that need to be held, they will turn to "defensive programming", vigorous testing, auditing, or just wishful thinking. There's another way! Enforce it using types! You will sleep better at night, trust me.
1
1
7
@PapiniShahar
Shahar Papini
3 months
The silliest guy I know is me from the past. What was this guy thinking? How come he didn't know what I know now??
@josephdelong
joseph.eth
3 months
“If you don’t wake up everyday and feel stupid you’re probably doing something wrong” Extraordinary advice from an investor
12
5
61
0
0
7
@PapiniShahar
Shahar Papini
9 months
A friend enlightened me that not everyone is aware of the ways Hamas incites hatred. A review of hate incitement in the schools of Gaza:
0
0
6
@PapiniShahar
Shahar Papini
3 years
@FroggyFrogster @ukolodny @LanrayIge @epolynya A single, yes. But composability between rollups is definitely possible
0
0
7
@PapiniShahar
Shahar Papini
1 year
@ThePrimeagen So, your first project is science, but your second is a lie?
0
0
7
@PapiniShahar
Shahar Papini
15 days
Personal take on building a dinosaur. Connie wanted a dinosaur. She asked both Sam and Genie to build one. Sam built legs, arms, tail and a body. Connie had to connect them herself. Genie went the extra mile and made an entire dinosaur figure. Next day, Connie wanted the
1
1
8
@PapiniShahar
Shahar Papini
2 years
@cronokirby The way we use STARKs at Starkware is very general. We use a lot of polynomial tricks. I think PLONK constraints are just a special case of what we consider AIR. (E.g. we use a mechanism similar to plookup)
1
0
7
@PapiniShahar
Shahar Papini
4 months
@gakonst We had to answer this very question on Starknet. Our answer was provably compiling to infallible native code that also computes gas (optimized) on both success and error branches.
1
0
7
@PapiniShahar
Shahar Papini
5 months
@RiscZero Wonderful work! But can it run crysis?? 😛
1
0
7
@PapiniShahar
Shahar Papini
5 months
@federicocarrone I played a ton of MedianXL back in the days...
1
1
7
@PapiniShahar
Shahar Papini
4 years
@ChristianPeel @StarkWareLtd @reddit @deversifi @Immutable @dydxprotocol We are exploring a couple of ways to compile from higher level languages (e.g. Haskell). Although, we will be happy to let the community help with building such tools:)
1
1
7
@PapiniShahar
Shahar Papini
1 year
@cairopractice @_tserg Anyone can do it, really:) Just make PR that adds these in corelib. God, open source is fun.
0
0
6
@PapiniShahar
Shahar Papini
2 months
@rel_zeta_tech @xyz_pierre Admittedly, I've been banging my head for hours trying to wrap my head around this diagram. I fear I will only truly understand it once I implement it.
2
0
5
@PapiniShahar
Shahar Papini
2 years
Another drawback: If we wanted to allow querying the state in the middle of a block, the API in incapable of expressing that.
3
0
6
@PapiniShahar
Shahar Papini
1 year
@GuthL @valkenburgh You are supposed to be on social media rehab!!! Let go of that phone!
1
0
6
@PapiniShahar
Shahar Papini
4 years
@VitalikButerin @loopringorg You can get away with keeping registration offchain. But for depositing actual assets (e.g. ERC20 tokens), you still need an onchain tx, right?
1
0
6
@PapiniShahar
Shahar Papini
1 month
@Rainmaker1973 I think these are stationary solutions, not stable ones.
0
0
5
@PapiniShahar
Shahar Papini
2 years
I believe the solution to this is introducing the concept of state_identifier, functions like getBalance and call should use it instead. The state_id itself could be specified in several way, for example StateAfterBlock(block_number), which would support the behaviour today.
4
0
6
@PapiniShahar
Shahar Papini
1 year
@raphael_dkhn If you use main, you can use loops
2
0
6
@PapiniShahar
Shahar Papini
2 years
There should be two concepts: a block identifier, and a state identifier. If we consider 2 blocks, B0 and B1, then we have *3* states we can refer to: S0 -> S1 -> S2 B0 B1 The API does not support querying at S0 (!)
4
0
6
@PapiniShahar
Shahar Papini
4 years
@jadler0 @StarkWareLtd @hkalodner @PhABCD @reddit @ethereum I wasn't clear enough, sorry. 32+1+8 ~ 42 B per change. There is 1 change per 3 txs. 42/3=14
1
0
6