Benchmarked 8 runs of @GoogleDeepMind Gemma 4 across quantizations, loaders, and speculative decoding setups. One prompt: explain ML-KEM with worked math.
โข 31B-4bit (lm) โ best accuracy, 6.8 tok/s, 18.4 GB. Still the winner
โข Speculative decoding hits 11.9 tok/s but degrades output quality
โข Matched QAT draft (72% accept rate) > mismatched (53%), but inherits weaker base model
โข VLM loader = same accuracy, 1 GB, no thought-chain
โข More bits โ better output. QAT โ better output.
The boring standard 4bit quant keeps winning.
โWhatโs in crypto for me?โ
Itโs the question I get asked the most at events as a speaker.
Instead of answering each time I decided to centralise the request:
For today's Bitcoin Pizza Day event I built YCrypto (WIP).
Just paste your domain and in 30 seconds it shows you the real crypto opportunities for your business.
No registration. No hype. No fluff.
โ Try it here: ycrypto.site/
Why this, Why now and How.
I have had my questions, so I hadn't said anything yet. But got it all clarified now, and I'll say it,
STRK20 IS A THING OF BEAUTY
Seamless network state integration and native access to network wide AA inside a ZK STARK is an absolute [W]
strk20.starknet.io
Nice work @Starknet@StarknetFndn@StarkWareLtd teams.
Say hello to the Cairo Debugger! ๐๐
We're excited to introduce our new project bringing ๐๐ถ๐๐ฒ ๐๐ฒ๐ฏ๐๐ด๐ด๐ถ๐ป๐ด ๐ถ๐ป ๐๐๐ to the Cairo ecosystem ๐๐จ
Cairo Debugger is integrated with the newest snforge from Starknet Foundry 0.60.0 ๐
๐ Read the docs & get started: docs.swmansion.com/cairo-debโฆ
Feeling a little dreamy but is this stupid?
When you compile this, your Starknet contract expects a proof here, verifier automatically generated for the described circuit.
Seamless privacy embedded within Cairo contracts.
1/ Starknet now has the native infrastructure for privacy.
The Shinobiโs upgrade is now live on mainnet, introducing the protocol-level changes that make privacy natively possible on Starknet ๐งต
x.com/StarkWareLtd/status/20โฆ
The sexiest Starknet update ever - Starknet v0.14.2!
So excited about fractal scaling with @ohayo_dojo Katana!
And also cheap network native ZKP โค๏ธ
governance.starknet.io/#/sn:โฆ
Thanks @Buremo_irede for the awesome demo! ๐ฅ
MISTzap = seamless multi-token private payments via QR codes on @Starknet โ completely gasless.
Cross-token privacy done right.
App: mistzap.vercel.app
Who else is sending crypto privately? ๐
Day 11 testing a Starknet app built with Starkzap
Here is MISTzap, a simple way to send crypto privately using QR codes, completely gasless. You can easily send and receive funds, and itโs already live on Starknet mainnet.
If privacy matters when sending crypto, MISTzap is worth checking out. Built by @_mistcash.
Link:mistzap.vercel.app#30DaysTestingAppsBuiltWithStarkzap
Day 11 testing a Starknet app built with Starkzap
Here is MISTzap, a simple way to send crypto privately using QR codes, completely gasless. You can easily send and receive funds, and itโs already live on Starknet mainnet.
If privacy matters when sending crypto, MISTzap is worth checking out. Built by @_mistcash.
Link:mistzap.vercel.app#30DaysTestingAppsBuiltWithStarkzap
Day 10 testing a Starknet app built with Starkzap.
Here is Gift app for sending gifts to friends and family. You can also stake your tokens to earn huge yield. No wallet needed, just your email and youโre good to go.
This great app is built by the king himself, @akashneelesh
Link: giftapp-starkzap.vercel.app/#30DaysTestingAppsBuiltWithStarkzap
Quick announcement: After long and heavy suffering :) the S-two white paper is finally out:
eprint.iacr.org/2026/532
Although nothing new in regard to the basic principles (a circle STARK, etc.) the white paper yet contains several details of broader interest:
- A formal description of the flat AIR circuit model (used by several contemporary zkVMs)
- A thorough soundness analysis of multi-table proofs: If one does not use "lifted" FRI, taming the soundness error turns out to be more sophisticated as expected. We introduce the notion of "cross-domain correlated agreement", and show that multi-table FRI satisfies this property.
- A discussion of adjusted conjectures, which takes into account the recent boost of proximity gaps counter examples. We believe that it is plausible to hope for acceptable list- and line-decodability properties up to the information-theoretic barrier, the Elias bound.
Thanks for all the help from the StarkWare team, and in particular to Dmitry Krachun for the many helpful discussions around his counter example.
Blockchain is financial freedom.
You can send USDC to a friend directly.
No banks. No intermediaries.
But...
The moment you send it, your friend has your wallet address - and with it, your entire financial history.
That's not financial freedom.
Blockchains need privacy.
Our MISTzap submission for โก๏ธStarkzap Blitz.
MISTzap allows sending private blockchain payments with QR codes.
Try it live at mistzap.vercel.app
PS, withdrawal UX is not developed yet, so please test with small amounts.