Gergely Orosz

Categories within Brand Verticals that narrow down the scope of Brands

Business Taxonomy

Brands, companies, advertisers and every non-person handle with the profit intent related to softwares, apps, communication equipments, hardwares

Technology Business

Brand

OpenAI

Products created by Brands.  Examples: Ford Explorer, Apple iPhone.

Product

ChatGPT

Unified Twitter Taxonomy

Technology

Artificial intelligence

Blogging

Writing

for individual and types of technology, e.g., food technology, 3D printing

After having read it, I can say this is probably the best book to explain how ChatGPT (and LLMs) work (written by Stephen Wolfram, who excels at explaining complex topics simply, so perhaps not a surprise)

There's also a blog post form for those wanting to read online. https://t.co/UsRftnc7ge

Really like how accessible but also thoughtful this book is.

Instead of jumping into the mechanics of ChatGPT (that it gets to) the first part of the book is about the evolution of neural nets (and deep learning), limitations and the concept of computational irreducibility. https://t.co/gPIvaz7wcf

Writing @Pragmatic_Eng, the #1 technology newsletter on Substack. Author of @EngGuidebook. Formerly Uber & Skype.

Stephen Wolfram explores the broader picture of what's going on inside ChatGPT and why it produces meaningful text. Discusses models, training neural nets, embeddings, tokens, transformers, language syntax.

What Is ChatGPT Doing … and Why Does It Work?

For those of us who have not spent the past few years building ChatGPT from the ground up, how does it work? From Evan Morikawa, who leads the Applied engineering team at OpenAI

How does ChatGPT work? As explained by the ChatGPT team.

The interactive blog post form.

If you do not fully understand how ChatGPT / LLMs work: spend 30 minutes on this and you'll understand how none of it is magic. https://t.co/2lfeYfelx0

Also, an explanation from the ChatGPT team: https://t.co/3udkZs5Jcm

Every time there’s a market decline or a market up-or-down:

I remember how when I worked at JP Morgan, these were times when traffic spiked and we made 5-10x the usual profits.

The safest place to be on any volatile market is being the middleman, taking a cut of every buy&sell.

2010 summer Greece was about to get bailed out (or default), JP Morgan predicted volatility for months.

So we froze all deploys, did nothing for 4 weeks but refresh daily stats.

Clients both won big and went bankrupt: being in the middle was extremely boring, but predictable.

“Why are we not building stuff??”

“Because we make far more money by doing thing. Wait and you’ll see.”

I was early on in my career at this time that doing nothing drove me crazy. It still would today, even if doing nothing is the business sensible thing to do.

Writing @Pragmatic_Eng & @EngGuidebook. Advisor @mobile__dev. Uber & Skype alumni. Career opportunities for software engineers & EMs: https://t.co/MJ0eAtlaS1

Remote work is not only creating more opportunities+better earning potential for many devs. It’s also turning software engineering into a much more competitive field.

Getting a a full-remote, highly paid/prestigious position means competing with people from around the world.

Hiring full-remote globally (eg using Remote, Deel or similar) increases the hiring pool by 5-20x for companies.

In practice, this means 5-20x more options for those perceived as the very top of the market. For many others, it means more competition, and higher expectations.

Full-remote work and global teams are also a fantastic “equalizer” in skills+confidence

Before I worked at Uber with SV teams: I always assumed those folks are better/ smarter than other regions.

Working together helped both “level up” & also understand what the global bar is.

An example to illustrate “globally competitive”.

@Replit is one of the hottest startups. Incredible growth, small but mighty team (~40 ppl), compensation that rivals Big Tech (and surpasses it, if the growth continues).

To get in, you’re competing with these types of folks: https://t.co/IiB6GCMiNb

Rest of World

Founders outside the US trying to hire/retain sw engineers: "It's a bloodbath. We cannot compete."

Engineers in lower cost regions: "This is the best time to be a developer. We are basically like rockstars.”

@restofworld
 writes about the hot market I've also been covering. https://t.co/uNwcKlvlOa

Full article: https://t.co/9pH2wEoFoq

The same is happening in places like Canada, Europe, Asia, based on founders I've talked with. It's a very good time for engineers with experience.

I'll be covering more details in tomorrow's newsletter.

I wrote about the hot market, how it compares to the Dotcom Boom, and advice to make the most of it here: https://t.co/itIS5WVFQk

If you're a hiring manager, details on what's going on, why, and what you can do: https://t.co/4pSmBEoQmE

Thinking back how much time every year me and other managers would spend arguing on whether X or Y should be in “strongly exceeds” calibration bucket or not.

And how the difference in bonus would still be minimal in the end, and people getting the most sometimes still unhappy.

While it’s easy to throw dirt on any performance process (especially from the outside): the reality is that any approach you choose will have downsides, and people unhappy about it.

The performance review process and culture+incentives go hand-in-hand and need to be deliberate.

Two extreme performance review processes:

1. Netflix. No performance review. No bonuses. Top of market comp. But fail the Keeper Test and you’re fired.

2. Uber up to ‘16. Stack ranking with extreme rewards for the top 0.5-1% at the top, at 5/5 (~10x target bonus).

I remember doing the math at Uber: the endpoint my team owned on a critical path had ~$144K/minute in revenue going through it, on average. So a 20-minute outage would be a cool ~$2-3M outage.

Yes, it did go down during my time. I started to appreciate blameless postmortems! https://t.co/XCK2GacfQ8

I’ll have to write about the oncall hell on my team which the team’s manager should have fixed so much earlier.

Oh, wait. I was that manager 😞

Until then, here are good practices from the industry for incident management I wrote up: https://t.co/55KOCOM4cm

This has been sitting in my drafts since I left Uber. I’ll get around to publishing it. Lots of lessons learned the hard way. https://t.co/dZ4dCaN2U8

"I worked at Big Tech for two years. Got promoted from Eng1 eng to Eng2 in a year. When I was not up for senior promotion, I left for a senior title at a small company.

I now fail to get senior offers at the rest of Big Tech and realize I had a great manager."

You learned.

"What advice do you have?"

Not what you want to hear.

Decide what you really want. Title or to learn?

Go to a place where you can see yourself learning and staying for 2-3 years. You have plenty of options.

You're only 2 years into a potentially 40-year career.

And here's an unconventional option that I have seen work well:

Go back to the first company to that great manager you had. I assume you left on good terms.

Works more often than you think.

I remember I being somewhat frustrated years ago how my side projects did not help me get jobs, weren't recognized during perf reviews: they were invisible for my corporate career.

They did set the basis for the entrepreneurial path I am on. They were an investment in myself. https://t.co/VwJhymbGmC

I only understood the "why" when I was an eng manager, talking with a dev who reminded me of my earlier self. I said:

"Know that this is a company who cares about what you do *for the company*. If they help the company: you get recognition. But do keep doing these for yourself."

When you are an employee, your employer recognizes and rewards things that directly affect their business.

If they gave you a higher bonus for some non-work-related success, they'd incentivize everyone to do stuff outside work, that the company sees no direct benefit from.

Also, some companies and managers will be hostile for too many, or too visible side projects.

I was told to do fewer talks (outside of work), blog less (outside of work) etc. I stood out from my peers with these activities, my (skip) manager noticed this, and did not like it.

In some ways, I now understand the hostility or not appreciating side projects too much.

Those who do many side projects are entrepreneurial, and there's good chance they won't end up the corporate career ladder. It's a pattern some of my former (skip) managers might have seen.

One of the things working at Uber has taught me is appreciating the non-tech folks at a tech startup.

As much as I'd like to think that technology made Uber huge, in reality, just as much (if not more) of a success came from things outside tech: making things work on the ground.

Uber had city launcher teams who did incredible work with whatever resources they had.

The "ops" people would build DIY tools, and hack around things when the engineering team had no bandwidth to build.

E.g. we shipped iDEAL on Rides with zero new code, all powered by "hacks".

I now see some early-stage startups hire operations folks who do exactly this.

They sort things for customers, hacking around, doing manual work, using Google Sheets, Zapier etc.

When the work is too manual *and it's needed* only then do these teams build automated solutions.

You'd think that doing things manually is inefficient.

On the contrary: it helps teams focus on what *really really* needs to be built.

As engineers, we tend to not admit the truth: the most expensive solution to any problem is to build it yourself, from scratch.

Another thing I remember is "faking" things to see if there's customer demand.

Instead of building the feature to encourage people to order quickly so we can batch, we just shipped a counter that gave $5 discount if you ordered within 5 minutes.

Then observed what happened.

Yesterday the CPO at Checkout .com (valued $40B) and the founder at Mollie ($6B) reacted to my claims that not all engineers get equity at these companies.

My take is simple. You can't call yourself a *tech* startup/scaleup if *all* engineers don't get *some* equity.

It's fine to disagree with this: it's my take. Ideally: equity should be for all employees.

It's how all companies *I* refer to *tech* scaleups and startups operate.

Equity allocation at private companies is almost always with a 1-year cliff: you need to last this long.

And don't get me wrong, companies like Mollie and Checkout I hear good things about engineering cultures, the people hired etc.

Right now, equity is a gray zone. You might get it. Maybe not.

Tech startups issue this to everyone who performs at the level they are expected to.

There is a reason the first question on The Pragmatic Engineer Test is equity-related: https://t.co/RK7HIKf5Pl

I understand some founders will disagree, want to use equity more as a reward to not all engineers, but only standout ones.

I'd not call these places a "startup".

The reactions:

Checkout dot com did not say all employees or all engineers have equity. But it seems there's progress.

Mollie seems to allocate equity later (probably?). The founder and a current engineer jumped in. Seems like if you're good, you'll (hopefully) get something. https://t.co/kmH6DFNnjP

@GergelyOrosz's Latest X Threads

Read threads by Gergely Orosz and get notified about future ones.