Parviz

An absolute geek. Almost anything in tech gets me excited. AI/ML & data get me even more excited! Views expressed are my own

[Parviz @pdeyhim]

@jbulltard1 Thank you!

[scott belsky @scottbelsky]

https://t.co/1fnS4oxAz3

[sysls @systematicls]

https://t.co/FQe5bCBqW1

[Zach Mueller @ CVPR @TheZachMueller]

A little birdie dropped by to give me the beast today. This thing is MASSIVE. Thanks @cfregly 🫡

[Bourbon Capital @BourbonCap]

Alibaba $BABA Chairman Joe Tsai: "We developed cloud computing really out of necessity, out of the need to become self-reliant"

"If we continue to use third-party software and hardware, we will later on hand over all of our profits to these technology vendors"

I absolutely love Alibaba’s mindset. Just a few companies out there are trying to remove the middlemen and trying to be self-reliant.

The company went from having excessive technology vendors eating all their profits to developing their own cloud computing, then to building their own chips, data centers, drones, AI models, robots, etc. Soon they will have their own planes, power plants, vans, cargo ships…

It’s the same vibe you get from $GOOG and $AMZN, working every single day to completely eliminate the middlemen and trying to be more self-reliant.

[Tim @timgedenk]

The death of the average.

We are currently witnessing the total collapse of the marginal cost of creation. Copywriting, design, video editing - skills that previously commanded a premium due to the barrier of technical execution are being democratised to the point of irrelevance.

Most marketers view this through the lens of efficiency. They see a tool that allows them to produce 10x the output for 1/10th of the cost.

When the supply of "good enough" content becomes infinite, the economic value of that content plummets to zero. We are entering an era of infinite noise. If you think it is hard to capture attention now, wait until the internet is flooded with billions of synthetically generated articles, tweets, and videos every single day.

(Which is already happening, just not at the quality and volume that it will in 6, 12 months from now)

In this environment, pure volume is no longer a valid strategy. You cannot out-publish a server farm.

The alpha in modern marketing is shifting entirely from production to provenance.

[1] The Trust Premium

As the internet becomes increasingly synthetic, we will see a massive flight to safety. "Is this real?" will become the single most important buying criterion.

We are moving away from algorithm-optimisation and back towards human-optimisation. Personal brands, founders-led sales, and verified human voices will command an exorbitant premium.

The faceless corporate brand is dead. If a consumer cannot verify the human source behind the message, they will subconsciously label it as "spam".

[2] High-Friction Marketing

For the last decade, the goal was "low friction". SEO, programmatic ads, automated email sequences.

As AI cannibalises these low-friction channels, - bots clicking on ads served by bots on sites written by bots - the smart money will move to high-friction channels.

Live events. Physical mail. Handshakes. Closed-door dinners.

The harder it is to scale, the more valuable it becomes. You prove your value by doing things that cannot be automated.

[3] Taste as a Moat

Large Language Models function by predicting the next most likely token. By definition, they regress to the mean. They give you the average of the entire internet.

If you use AI to guide your strategy, you are opting for mediocrity at scale.

"Taste" - the human ability to curate, to select the outlier, to understand nuance and subtext - becomes the only defensible moat.

The future of marketing is about who has the taste to know what *not* to create.

Paradoxically, the more artificial the world becomes, the higher the premium on being undeniably human.

[mitsuri @0xmitsurii]

Remember the time when Steve Jobs pulls a MacBook Air out of an envelope and made marketing history.

[Angry Tom @AngryTomtweets]

no words

[SemiAnalysis @SemiAnalysis_]

While you doing gaming gpus & rushing to add DP4A & WMMA, I studied all the limits of physics.

While you were out in an marketed theoretical flop race with AMD, I mastered jax xla<>TPU hardware software codesign to maximize MFU.

While you wasted your days adding pyxis plugin to slurm, I cultivated xmanager.

And now that the world is on fire and the barbarians are at the gate, you have the audacity to beg me for help?

[Steve Burns @SJosephBurns]

"Intelligence consists in ignoring things that are irrelevant." - Nassim Nicholas Taleb

[Andrej Karpathy @karpathy]

I put up a simple repo I call reader3 (it's my 3rd version...) to illustrate how I read EPUBs with LLMs. Basically get some epub (e.g. Project Gutenberg is great), go chapter by chapter, and with this you can easily copy paste text to your favorite LLM.
https://t.co/HoBDxCJhdC

[Andrei Verdes @theandreiverdes]

My coworker got promoted over me.

He was worse at coding. Better at politics.

I wrote better code. Fixed more bugs. Shipped faster.

He talked in meetings. Took credit. Played the game.

He got the promotion. I got "keep up the good work".

That's when I realized: corporate rewards politics, not performance.

Six months later I quit. Started freelancing.

Now I make 3x his salary. No politics. No credit-stealing. Just solving problems and getting paid directly.

The best developers rarely get promoted. They get used.

Companies optimize for compliance and communication. Not competence.

If you're technically great but politically terrible, you'll never win at corporations.

Leave. Build your own thing. Get paid for your actual value.

[Justin Welsh @thejustinwelsh]

Your employer will never pay you what you're worth. They can't.

Their model only works if there's a gap between what you produce and what you cost.

A great way to close that gap is to work for yourself.

[Divya Mangotra Manchanda @DivyaMangotra]

The ultimate form of preparation is not planning for a specific scenario, but a mindset that can handle uncertainty.

— James Clear

[Ray Dalio @RayDalio]

Remember this: The pain is all in your head. If you want to evolve, you need to go where the problems and the pain are. By confronting the pain, you will see more clearly the paradoxes and problems you face. Reflecting on them and resolving them will give you wisdom. The harder the pain and the challenge, the better.

Because these moments of pain are so important, you shouldn't rush through them. Stay in them and explore them so you can build a foundation for improvement. Embracing your failures--and confronting the pain they cause you and others--is the first step toward genuine improvement; it is why confession precedes forgiveness in many societies. Psychologists call this "hitting bottom." #principleoftheday

[GLADIA Research Lab @GladiaLab]

LLMs are injective and invertible.

In our new paper, we show that different prompts always map to different embeddings, and this property can be used to recover input tokens from individual embeddings in latent space.

(1/6)

[Nicholas Fabiano, MD @NTFabiano]

Instead of “I’m bad at this”.

Say “This is new for me”.

This gives your brain the space to learn instead of shut down.

This is neuroplasticity in real time.

[Paulo Macro @PauloMacro]

Sounds like a great idea, real innovation…

[Parviz @pdeyhim]

https://t.co/bEbkRJF9Cz

[Nassim Nicholas Taleb @nntaleb]

I just published The World in Which We Live https://t.co/wNWwFJsIIn