Klarna was the company that went all-on replacing customer support with an AI bot and went on to brag about the cost savings.

Now they are reversing course.

Easy to see more companies blindly replacing quality customer support with a worse AI implementation will follow... Back when Twitter was full of influencers declaring the end of customer support thanks to Klarna I did something few people did:

Signed up for Klarna, bought an item, and used the bot.

I was NOT impressed. At all.

Called that this was... very basic. blog.pragmaticengineer.com/klarnas-ai-cha…

"Leetcode-style / DSA / algorithmical interviews are useless and don't measure what's really expected on the job. They are also inefficient, and companies using these are hiring for the wrong people."

Heard this SO many times.

The responses almost always miss the point.

I'll do a longer post one day, but a few thoughts:

1. YOU are not Big Tech. You probably don't have 1,000+ qualified applicants show up for an entry-level job posting and 100+ for a senior posting - in just a day or two, without advertising it

2. When a company gets large enough combined with #1, the game becomes not reducing false negatives but reducing false positives to zero

3. "LeetCode-style interviews are BS and don't measure what you do on the job." Yes. This is part of the reason. Guess what else is BS at Big Tech? A lot of stuff? Do you think people who are unwilling to put up with BS (that has historic context and can be internalized) would last at these companies? No: they would quit shortly or be pushed out as they refuse to do what everyone else does. These interviews conveniently self-select for people who can and do put up with BS

4. Career ladders. There is a notion that a Principal engineer should be as good or better than a new grad in every area - including algo coding. Like it or not, it's how it is

5. Technical managers. Many of these companies expect managers to pass the same bar. Like it or not, again: the reality is at these places many (probably all) line managers can code, and can do it very well.

6. Scalability of process. Have you ever had the challenge of onboarding 120 new interviewers in a month? Every quarter? These companies have this problem.

7. If it ain't broken: don't fix it.

Look at the business results of Big Tech. If the interview process would be broken, it would show up in eg shipping slower and being outcompeted by competition etc. In reality: Big Tech is more nimble than ever. E.g. Threads, Copilot, Gemini etc. Their interview process works *for them*

8. You are probably not Big Tech and don't have to solve for this very distinct set of problems. Remind me how Big Tech hiring is broken when they built a new social media network in 6 months from idea to launch. This was 2x faster than e.g. Bluesky (a nimble and amazing startup btw)

Threads story: newsletter.pragmaticengineer.com/p/building-the…

Bluesky story: newsletter.pragmaticengineer.com/p/bluesky

Via @Pragmatic_Eng

I rarely do predictions but this is an easy enough one to make:

Big Tech will bring back onsite interviews for the final round of interviews, flying in candidates.

When you pay top of market, AND do hybrid work, in the age of invisible AI helpers, remote interviews are a risk

https://twitter.com/tom_doerr/status/1890842725655658641

A company expecting staff to work in-office 2-3 days per week will increasingly prefer in-person (final round) interviews.

If they pay top of market: this itself will be enough for most candidates to do it. The payoff is high enough, after all.

In-person interviews also negate all "cheating" that can be done with AI. It also means existing interview formats (eg algo interview, sytems design etc) don't need to be changed to remain as effective as before!

Don’t fall for speculation like this.

Software developer job postings are back to where they were in ~2019, following the COVID boom (and zero interest rates!).

You know what else has the same pattern? Banking jobs. Marketing jobs. Heck, HR jobs:

https://twitter.com/gregisenberg/status/1889847881424646403

Banking and finance jobs:

The story is not how software developer jobs are disappearing because of eg AI.

The story is how there was an economic boom 2021-2022, massive hiring, and now a correction.

A lot to do with zero interest rates! More: newsletter.pragmaticengineer.com/p/zirp

Building a profitable company as a commercial open source one with a permissive license feels more challenging these days.

The company needs to make money: which it usually does by selling a managed version of the software.

But here’s the catch: (cont’d) 1. If the open source version of the software is standout and easy to operate: users have little to no incentive to choose the managed version

2. If the license is restrictive, competitors can simply undercut pricing. They don’t need to invest in development of the software much

Unpopular opinion:

Developers are some of the most demanding customers to please with developer tools, and are one of the hardest business to make a profit in.

When we feel something is a ripoff... we'll build ourselves/migrate/adopt a new tool.

Often out of spite! Can't really point to another business where customers will sometimes leave a vendor even if it costs them a *lot more* to migrate off, or build a tool.

Unsaid is how it's professionally satisfying to prove that we can build a complex tool, plus it's often rewarded by leadership as well!

The more I use GenAI coding tools, the more I am convinced keeping to "traditional" software engineering practices is what works most productive here. As in 10x more productive. E.g.

- Small changes
- Test that the change works before moving on
- (unit) tests wherever you can Amusing how error-prone all of these models are

I catch how often it generates buggy code thanks to testing and adding unit tests (I have it usually generate tests and then I tweak for my test cases)

Don't know how people are productive who let it run loose tbh

One interesting I’m seeing with GenAI coding tools:

The MASSIVELY help technical founders at small and mid-sized startups prototype, challenge dev team, and ship products faster.

A recent example I’ve seen (cont’d): Founder: “here’s a product idea we should do.”

Dev team: “Ok. We’ll build a prototype. It will be ~2 weeks.” 1 month later there’s a prototype. Another 2 months to ship to customers.

Now: founder builds prototype in ~4 hours, shows to dev team. Team builds a more prod ready one in a week and ships to customers!

YouTube helpfully shows what other videos people watch who stop by at The Pragmatic Engineer Podcast YouTube channel

This is either a bug, or @ThePrimeagen managed to hack YouTube (not implausible) or he went back in time to upload a video in 1970

Or.... OH WAIT I KNOW If you know why this says 55 years ago don't say a thing!!

(if you know you know)

Interesting:

Meta created React Native. It’s used (with components at least) in their flagship apps: Facebook (iOS, Android), Instagram (Meta Quest), Messenger (desktop).

Google created Flutter. And yet none of their flagship apps use it (Gmail, YouTube, Maps, Calendar). The only flagship Meta app not using React Native is WhatsApp.

Google does build a lot of smaller apps with Flutter.

Just odd that Flutter can be used as modules (for a few screens) but Google, for some reason, doesn’t do with major apps.

Food for thought.

Record revenue (on track for $156B), record profits (on track for $55B), fresh enough memories from firing 25% of staff... and Meta is at it, again.

Ruthless business rationale, as the goal is not cost savings, but firing the old, hiring the new.

Good luck to everyone at Meta. Source:

Hard to believe that ~2 years ago, Meta had not done any mass layoffs throughout its entire history (2004-2022)

Now it looks to become an annual tradition. What a change.bloomberg.com/news/articles/…

This release is a big deal because:

1. It's the second major agentic coding workflow released with public access. (First one was Devin)

2. Pricing. Unlike Devin that charges $500/month right now. Copilot Workspace Preview is free.

3. GH integration. A big sell+onboarding win!

https://twitter.com/satyanadella/status/1878578314115473577

This is how Copilot Workspace works. Covered 8 months ago in @Pragmatic_Eng at

I personally think it's a clever workflow that aims to make devs more productive (and not replace them, like tools like Devin advertise themselves to the business) newsletter.pragmaticengineer.com/p/the-pulse-92

Fascinating to see how different games development is to e.g. SaaS development (or anything backend or frontend development).

"Job well done" means something very different in both cases. And both dev lifecycle and what "great" means is very different between one another. This reflection comes after talking with best-selling game developer @JonasTyroller and asking if they do unit tests or code reviews.

They do neither.

It's not what makes a smash hit game with a tiny team (that his latest game is!)

What happens if a company’s engineering team suddenly becomes a LOT more productive: more story points shipped, self-reported 20-50% productivity increases after using GenAI tools (that the company pays for.) No HC change.

BUT

Revenue stays FLAT despite all this

What then? Hints as to why revenue might stay flat:

Competitors are doing the same and their productivity increased by just as much. And customers don’t care about how you do against your old self; they care about how you do against competition!

Imagine having a $300K+ year job at Apple. Then notice Apple matches charity donations up to $10K/year and deciding… you want that additional $10K/yr. So you defraud Apple with a fake charity to funnel the matching contribution back to you.

6 ex-employees did this. Just… wow: Full article and details:

What a way to throw away your corporate career, for practically nothing in the grand scheme of things.

I assumed someone smart enough to get into Apple would not be dumb enough to commit such fraud.nbclosangeles.com/news/national-…

Talk about a board turning a company into nothing.

Bench raised a $60M Series C in 2021 - then board pushed out the founder CEO, to bring in a professional CEO.

The company is shutting down 3 years later. Not even a fire sale.

Customers, employees, investors left with nothing.

https://twitter.com/patio11/status/1872746338351497313

A rare glimpse of how brutal internal politics within a scaleup can be.

In the end, investors control the board. Board members can override the founder if investors get more power than the founder.

Bench could be a warning example why this is a terrible idea - for investors!!

I see so much FUD (fear, uncertainty, doubt) about the future of sw engineering, mostly from non-devs. Along the lines of “soon anyone can spin off AI agents in bulk that act as hundreds of devs.”

A false premise. Just open your airline app that is built by ~hundreds of devs. Good software is far more nuanced than just how many devs work on it.

Skype hired good devs and won the video calling desktop market. It got to 1,000+ devs and then… this startup with 50 devs washed the floor with them for mobile chat + calls. It was called WhatsApp.

An odd trend I’m noticing:

A lot more startups (esp AI startups) bragging about how much they work: in terms of well past midnight, 6-7 days per week, 12+ hour days etc.

What is the end game here?

Good job grinding. But it’s v hard to innovate + ship quality sleep deprived… Speaking for myself: yes, when I’m locked on a problem I’ll work more *to get it done.*

But when I regularly pull late nights, my work (+mood, judgement) gets worse, not better.

Best work is (and has been) usually after taking enough rest and being full of energy+motivation.

Automattic - the creator of WordPress, a company raising $950M in VC funding - took a paid WordPress plugin built and owned by another dev and re-published it, making it free.

If you have a business selling a paid WP plugin: Automattic can null it, anytime.

Another new low.

https://twitter.com/youngbloodjoe/status/1860559013563298182

A summary of past events:



I used to be a big Automattic / WordPress fan (my blog used WordPress for many years, and I admire companies investing in open source.)

Automattic has turned into a corporation ignoring open source ethics though, in its bid to take out its biggest rival WP Engine.blog.pragmaticengineer.com/did-automattic…

Worth asking: why would a web infra company acquire a code search startup?

Answer this question, and you get interesting insights on the strategy Vercel is likely pursuing.

https://twitter.com/rauchg/status/1859365672444363037

It's not a one-step answer.

Q: "What does a code search tool need?"

A: "Access to a customer's whole codebase to work well."

What is happening at Sonos is hard to believe.

The company released a new app that is a big regression in reliability, usability and functionality vs the old one.

Sales are falling thanks to the new app.

There is no quick fix.

They want to re-release the old app... but cannot Screenshot from

Talk about a company damaging themselves, thanks to ignoring practices like:

1. Only release a product that has been properly tested (was not the case w Sonos)

2. Have a rollback plan, especially when skipping #1 (also not the case)reddit.com/r/sonos/commen…