Shapor Naghibzadeh

Chronicle Co-Founder / AI Lead, Google Cloud Security

[Shapor Naghibzadeh @shapor]

🔗 6,576 days ago, I almost didn’t take the job → https://t.co/ir8UkL4TBk

I was skeptical—of Big Tech (which wasn’t even that big back then), of a search company evolving fast, and of whether it was for me.

What was supposed to be a 3-month contract has turned into an 18-year journey.

Funny how the biggest opportunities often start with a little skepticism. 🚀

[Shapor Naghibzadeh @shapor]

The U.S. government’s official spending database (https://t.co/hDwCOBmHx1) can’t properly filter recipients right now due to a “bug,” yet Booz Allen Hamilton is paid tens of millions of taxpayer dollars each year to provide hosting, data services, and support for the site. If nearly $10 million annually (with a potential total of $45.9 million) is on the line, you’d think we’d get a fully functional transparency tool! 🤦
https://t.co/hymoPXeMdJ

[Shapor Naghibzadeh @shapor]

@JosvdWest @mckaywrigley Very cool to see others with similar workflows. Not exactly Repo Prompt, but I built GitIQ for this: Enter prompt, select files, then it combines the files into a JSON request to the model and applies file updates to the local git repo. https://t.co/xnBfDjYlUt

[Shapor Naghibzadeh @shapor]

@halvarflake Hah, great minds think alike! I prompted an eerily similar tool (with a file picker) in Python—because, as much as I hate it, the LLMs seem to love it. And I completely agree, diffs are the *only* way that make sense for reviewing AI-generated code!

[Shapor Naghibzadeh @shapor]

@argvee Cheers! 🥂 Don't forget celebrate with a sweet treat on xmas eve as well! 🍫

[Shapor Naghibzadeh @shapor]

🚀 Exciting News: Duet AI in Security Operations Goes GA! 🚀

Thrilled to announce that our LLM-powered features in Chronicle are now in General Availability as Duet AI for Security Operations!

A big congratulations to our team for the exceptional work in achieving this milestone.

https://t.co/GwvZHoSayq

#DuetAI #AI #Chronicle #CyberSecurity

[Shapor Naghibzadeh @shapor]

Do you remember when you joined X? I do! #MyXAnniversary

[Shapor Naghibzadeh @shapor]

Today at RSA we announced what I've been working on recently - Chronicle's new AI capabilities! If you're in town and want to connect, please feel free to reach out. https://t.co/jekXh0DbME

[Andrej Karpathy @karpathy]

The analogy between GPTs of today to the CPUs of early days of computing are interesting. GPT is a funny kind of programmable text computer. Have to think through it more 🤔 but e.g.:

## Memory
GPT-4 RAM is ~log2(50K vocab size)*(32K context length)/(8 bits/byte) ~= 64kB, roughly a Commodore64. Just as then, optimizing this precious resource is critical.
GPT registers are the residual stream. There are d_model of them, e.g. GPT-3 has ~12K registers. VLIW architecture vibes.

## CPU
The LOAD instruction is the Attention mechanism, except it can address by both location and/or content.
The STORE instruction is forced every n_layer number of clock cycles.
The ALU are the MLPs + LayerNorms. Awkwardly, as their params are not shared across layers, the ALU changes at each clock cycle. Optionally the MLPs may also be interpreted as supporting a kind of fixed knowledge database lookup.
The programs always takes the form [[LOAD, ALU]*N, STORE]*M, where N is n_layer and M is num_tokens.

## Architecture
GPT feels closer to a fixed-function than stored-program computer because the number of parameters is so large. In contrast, the description length of a CPU is very low and all the action is in the memory configuration.
Another way to look at it is that GPT is a much more bloated/complex computer. Which is fine because it is not engineered but optimized and the upshot is that the programs can be shorter.

[Siemens USA @SiemensUSA]

#PressRelease: #Siemens and @chroniclesec will partner to provide industrial monitoring and detection for the energy industry. https://t.co/dfiY38fupP #SiemensInnovates

[Eric Schmidt @ericschmidt]

This product is phenomenal and every company will want this on their network to keep safe

[Shapor Naghibzadeh @shapor]

If you've been wondering what I've been up to the past few years - here it is!

[Google Cloud Security @GoogleCloudSec]

Today we launch #Backstory, the first global security telemetry platform designed for a world that thinks in petabytes. Come find us at #RSAC booth #2251 South after 5pm and learn more. https://t.co/Og5HZswDnV

[Shapor Naghibzadeh @shapor]

Alphabet's Moonshot Factory And Stephen Gillett Create Chronicle To Revolutionize CyberSecurity via @forbes https://t.co/aKFfok2YCW

[Stephen Gillett @stephengillett]

Proud to launch our new company! Introducing Chronicle, a new Alphabet business dedicated to cybersecurity: https://t.co/FUIqHZyNSP @chroniclesec

[Shapor Naghibzadeh @shapor]

Super excited to announce our new company! https://t.co/QGjIokkUfz

[Google Cloud Security @GoogleCloudSec]

Hello, World! Introducing Chronicle, Alphabet’s newest company. Get the story: https://t.co/foaAWSNeP4.

#cybersecurity

[Shapor Naghibzadeh @shapor]

install ghostery + enable click to play plugins in chrome + 'ln -s /dev/null ~/.macromedia' makes me feel a lot better about browsing

[Shapor Naghibzadeh @shapor]

Awesome to see #perfkit getting traction at LPC! Great work @vwduder ! http://audidude.com/?p=495

[Shapor Naghibzadeh @shapor]

if you care about your privacy online, download the ghostery plugin (for chrome or ff) www.ghostery.com

[Hacker News Bot @hackernewsbot]

Reddit's Server HTTP response header is is "'; DROP TABLE servertypes; --"... http://bit.ly/ciVtUl