David Vernet

Ring 0 Linux kernel hacker at Meta, scheduler-bug-adder, BPF standardization co-chair. If you want to know how it works, break it apart.

[David Vernet @byte_lab]

Something that occurred to me the other day is that AI has turned writing high-quality software from a P problem into an NP problem.

Writing high-quality software used to be (in a loop):
(1) Read and understand a large, complex, and probably messy codebase
(2) Formulate a problem
(3) Design a solution for it
(4) Implement a solution for it that fits well into existing contours
(5) Verify that solution, and goto (1-4) if more iteration is needed

This is the process now:
(1) Formulate a problem (and describing it to the AI)
(2) Verify the solution that the AI gives you, and goto (1) if more iteration is needed

Some people seem to think that (1) is the whole name of the game now. This is COMPLETELY missing the point of the tool, though. It's supposed to get rid of the polynomial-time part of the software-development lifecycle. It's not supposed to make it O(1), and pretty much all of the AI-slop trash I see go by has clearly never had a human actually look at what it's producing.

None of this is surprising I guess, but if you consider yourself to be a strong engineer and tend to hold a high bar, I think it's very important to bear this in mind. ""Prompt engineering" and "context window engineering" are important (those are encapsulated by (1) above), but you have to also do the verification and meaningfully look at what's coming out the back end of the AI to fully bake what it's giving you.

It's kind of baffling to me that people don't do this, because you can still 10-20x+ your productivity, _and_ produce even better-quality code than you were before. It would be like someone deciding to bake 1000 cookies to 60% done instead of baking 700 cookies to 99.99% of the way done, when before you used to bake a batch of 24 to 100% of the way done in the same amount of time.

[David Vernet @byte_lab]

Ironically, this account is the epitome of AI trash slop. Nvidia hasn’t done a large layoff since 2008. Why would they start laying people off now when their company is in the stratosphere and at the epicenter of the AI boom?

[Kalshi Finance @Kalshi_Finance]

**NVIDIA JUST POSTED $96.3 BILLION REVENUE WHILE SECRETLY GUTTING 4,200 SOFTWARE ENGINEERS AND REPLACING THEM WITH THEIR OWN AI CODING AGENTS**

Jensen Huang made $26.8 million last quarter while his own engineers built the tools that eliminated their jobs

The most fucking dystopian tech story of 2026

NVIDIA didn't announce layoffs. They announced "AI-Native Development Teams"

4,200 software engineers. The people who built CUDA. The architects behind their AI compiler stack. The teams that made NVIDIA the most valuable company on earth.

All "transitioned" to make room for AI agents running on their own H200 clusters

I'm hearing the AI agents are literally trained on the terminated engineers' own codebases

Sources saying the productivity metrics are "inhuman" - 18x faster code generation, zero bugs, 24/7 development cycles

One senior architect told me they were asked to train their replacement for 6 months before getting managed out

The AI system ships features faster than the human teams ever could

Management keeps calling it "the future of software development" while the people who built that future get escorted out by security

Stock hit $2,847 today. All-time high. Based on technology created by people they just fired.

Jensen's net worth jumped $4.2 billion this quarter alone while 4,200 families lost everything

The engineers who made AI possible just got replaced by AI

If you're a software engineer anywhere, you're looking at your own obituary

[David Vernet @byte_lab]

Does someone want to remind this cretin that a huge portion of the US military runs on an OS invented and given away for free by a Finnish guy who now resides and pays taxes in the US?

[Stephen Miller @StephenM]

Look at the rate of civilizational advancement from 1st flight to moon landing and consider where we’d be now had we not subsequently reoriented society around the idea that the entire world has a right to live here, vote here & receive legal, political and financial preference.

[David Vernet @byte_lab]

Finally got a chance to start reading @hahussain’s The Arab Case for Israel. Imagine my surprise when he describes attending a Yom Hazikarok celebration at B’nai Israel Congregation in Maryland, where I was Bar Mitzvah’d, and whose head Rabbi officiated my wedding.

Small world, and it was quite moving to learn that it played a small role in the journey of such an intellectual powerhouse of the Middle East.

[David Vernet @byte_lab]

CachyOS has an awesome community around perf. They were the first distro to embrace sched_ext by a country mile and are also very welcoming and collaborative. These numbers are impressive but not surprising.

[CachyOS @CachyOS]

CachyOS is 14% faster than Ubuntu/openSUSE on Intel Panther Lake - just from distro-level optimizations. That's a massive gain for "free" performance. 🚀

Thanks for the benchmark @phoronix

[David Vernet @byte_lab]

I wonder if this guy used AI to write this tear-jerker hypothetical.

AI is definitely going to be an apocalypse for the tech industry, but it’s not just new grads who are in danger. The people with ideas are going to be the ones who succeed. AI is an equalizer in that regard.

[Kalshi Finance @Kalshi_Finance]

Kid graduated with a 3.87 GPA from a decent state school three months ago

$191k in student debt between undergrad and a CS masters degree

Has applied to 1,340 entry-level positions since graduation

Zero offers. Twelve phone screens that all ended the same way.

"We've restructured our junior pipeline around AI-assisted development"

"Senior developers with Cursor can now handle what used to require a full team"

"We're looking for candidates who can hit the ground running with minimal supervision"

His girlfriend landed a consulting gig making $95k. She has a sociology degree.

His roommate from the dorms got hired at Goldman doing Excel modeling for $140k. Finance major with a 2.9 GPA.

Meanwhile he's watching YouTube tutorials on prompt engineering at 2 AM wondering if four years of algorithms and data structures was completely fucking pointless.

Student loan payments start in January.

His parents keep asking when he's going to "use that expensive degree"

The career center at his school is still sending emails about "abundant opportunities in tech"

Last week he applied to be a shift supervisor at Target

They told him he was overqualified

[David Vernet @byte_lab]

Basically, this claim that everyone should become memory experts sounds nice but is meaningless. You’re asking everyone to become a deep expert who knows how every byte of their software is used.

No, the correct solution is to *decrease* their cognitive load, not increase it.

[David Vernet @byte_lab]

Broad proclamations like this are much easier to make than solving the problems they’re pontificating about.

The fact is that everything has a cost, including memory efficiency. Here are a few things to bear in mind 🧵

[Chayenne Zhao @GenAI_is_real]

unpopular opinion: 16GB is plenty if software engineers actually cared about memory efficiency. chrome eating 4GB for 12 tabs is not a hardware problem its a software disgrace. docker consuming 2GB idle is not a feature its laziness. we live in an era where people optimize every single token to save $0.001 on API costs but happily ship electron apps that eat 500MB to display a todo list. if the industry treated RAM the way we treat inference compute - obsessively measuring every byte - 16GB would feel luxurious. the hardware isnt the problem, the software is @adxtyahq

[David Vernet @byte_lab]

The alternative is expecting everyone to be an expert which is never going to happen. It’s similar in spirit to expecting everyone to write perfect multithreaded code. It’s impossible. The next best thing is to write runtime frameworks that do it for users.

[David Vernet @byte_lab]

Appreciate the shout out! I should really dust off the blog and get it going again. Let me know if there are any topics you want to hear about. I was thinking something like, “How do things actually run on CPUs?”. Can go into instructions, schedulers, etc.

[kshitij vaze @VazeKshitij]

Yo. Ever wondered what the hell is computer memory? Here is a 4-part blog series by @byte_lab that talks about it.

And trust me, it is some of the best worded explanation that I have come across -
https://t.co/MbAsVVqmLf

https://t.co/M1FEWCsOkB

https://t.co/ADbheP4jJO

https://t.co/52KAnhRTqw

[David Vernet @byte_lab]

“So far Apple has yet to mitigate these issues even with being disclosed to the company last year.”

[Phoronix @phoronix]

Apple CPUs Affected By New SLAP & FLOP Side-Channel Attacks

https://t.co/c9qeewd0mM

[@IETF @ietf]

With the BPF Instruction Set Architecture (ISA) document now officially published as #RFC9669 read about the details the process of bringing the RFC document to fruition, and why it's important to standardize core components of the BPF ecosystem: https://t.co/F7snL3KuYA

[David Vernet @byte_lab]

I wonder if aliens in the sombrero galaxy realize that other aliens look at their entire galaxy and are like, “Lol that’s a sombrero”

[David Vernet @byte_lab]

The BPF Instruction Set Architecture RFC document is hot off the presses: https://t.co/WdruzDjV3F. Thanks to the @ietf, and to everyone who participated and helped make it happen!

[David Vernet @byte_lab]

Very interested and excited to hear more details. I always assumed it was impossible to trace an inlined function because the compiler can do basically anything with it, including e.g. eliding it completely. Daniel knows this though, so I’m keen to see what he’s cooked up!

[Daniel Xu @__dxu]

Really exciting progress on finally solving the inlined function problem for tracing. Stay tuned.

[David Vernet @byte_lab]

@Naccyde Si tu aimes vraiment la conférence, il est temps de lui rendre la pareille en faisant une présentation 😉