Gergely Orosz
@gergely.pragmaticengineer.com
Writing The Pragmatic Engineer (@pragmaticengineer.com), the #1 technology newsletter on Substack. Author of The Software Engineer's Guidebook (engguidebook.com). Formerly at Uber, Skype, Skyscanner. More at pragmaticengineer.com
Every single time an LLM hallucinates, I am grateful:
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
November 9, 2025 at 2:20 PM
Every single time an LLM hallucinates, I am grateful:
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
I am so annoyed at scare tactics as a dark pattern trying to get you to upgrade your SaaS subscription.
This is Google Docs doing so. I am already a paying customer - and no, I don't have "sensitive files" shared externally.
Such a cheap way to try to get more $$, Google
This is Google Docs doing so. I am already a paying customer - and no, I don't have "sensitive files" shared externally.
Such a cheap way to try to get more $$, Google
November 8, 2025 at 9:22 AM
I am so annoyed at scare tactics as a dark pattern trying to get you to upgrade your SaaS subscription.
This is Google Docs doing so. I am already a paying customer - and no, I don't have "sensitive files" shared externally.
Such a cheap way to try to get more $$, Google
This is Google Docs doing so. I am already a paying customer - and no, I don't have "sensitive files" shared externally.
Such a cheap way to try to get more $$, Google
A free, non-VC funded lifestyle business idea:
AI training / workshops for devs
Massive demand from more traditional companies that have budget for this and are seeing devs not really picking up these tools. They want to invest
Doesn’t exist as far as they are aware
AI training / workshops for devs
Massive demand from more traditional companies that have budget for this and are seeing devs not really picking up these tools. They want to invest
Doesn’t exist as far as they are aware
November 7, 2025 at 7:27 AM
A free, non-VC funded lifestyle business idea:
AI training / workshops for devs
Massive demand from more traditional companies that have budget for this and are seeing devs not really picking up these tools. They want to invest
Doesn’t exist as far as they are aware
AI training / workshops for devs
Massive demand from more traditional companies that have budget for this and are seeing devs not really picking up these tools. They want to invest
Doesn’t exist as far as they are aware
Chris Lattner is one of the most influential engineers of the past two decades. He created LLVM, Swift, contributed to TensorFlow, and created the Mojo language.
What was the story about creating Swift - and why did he face resistance inside Apple when wanting to replace Objective C?
(cont'd)
What was the story about creating Swift - and why did he face resistance inside Apple when wanting to replace Objective C?
(cont'd)
November 5, 2025 at 9:23 PM
Chris Lattner is one of the most influential engineers of the past two decades. He created LLVM, Swift, contributed to TensorFlow, and created the Mojo language.
What was the story about creating Swift - and why did he face resistance inside Apple when wanting to replace Objective C?
(cont'd)
What was the story about creating Swift - and why did he face resistance inside Apple when wanting to replace Objective C?
(cont'd)
Blind (and much of mainstream media!): the AWS outage is the result of “mass layoffs, outsourced talent.”
Me: talks to the AWS engineers handling the incident. Turns out the creators of the systems impacted were in the call (not laid off!), no outsourcing etc
Will share more
Me: talks to the AWS engineers handling the incident. Turns out the creators of the systems impacted were in the call (not laid off!), no outsourcing etc
Will share more
November 5, 2025 at 9:18 AM
Blind (and much of mainstream media!): the AWS outage is the result of “mass layoffs, outsourced talent.”
Me: talks to the AWS engineers handling the incident. Turns out the creators of the systems impacted were in the call (not laid off!), no outsourcing etc
Will share more
Me: talks to the AWS engineers handling the incident. Turns out the creators of the systems impacted were in the call (not laid off!), no outsourcing etc
Will share more
Coming tomorrow on @pragmaticengineer.com - with Chris Lattner!
November 4, 2025 at 9:18 PM
Coming tomorrow on @pragmaticengineer.com - with Chris Lattner!
The promise of AI chat assistants: they solve 90% of the problems users have (by looking up the docs and telling them)
My reality: need to spend 10 minutes trying to get to a human, to solve an issue I need customer support to look into
Around minute 8 I sign up to a competitor
My reality: need to spend 10 minutes trying to get to a human, to solve an issue I need customer support to look into
Around minute 8 I sign up to a competitor
November 4, 2025 at 4:14 PM
The promise of AI chat assistants: they solve 90% of the problems users have (by looking up the docs and telling them)
My reality: need to spend 10 minutes trying to get to a human, to solve an issue I need customer support to look into
Around minute 8 I sign up to a competitor
My reality: need to spend 10 minutes trying to get to a human, to solve an issue I need customer support to look into
Around minute 8 I sign up to a competitor
Rare to read a post arguing about the importance of software architecture BUT doing it without software architects than this one.
And you can just sense the hard-earned scars Matt got on the way: at Netflix, Twitter & other places:
Such a good read: newsletter.pragmaticengineer.com/p/what-is-go...
And you can just sense the hard-earned scars Matt got on the way: at Netflix, Twitter & other places:
Such a good read: newsletter.pragmaticengineer.com/p/what-is-go...
November 4, 2025 at 7:57 AM
Rare to read a post arguing about the importance of software architecture BUT doing it without software architects than this one.
And you can just sense the hard-earned scars Matt got on the way: at Netflix, Twitter & other places:
Such a good read: newsletter.pragmaticengineer.com/p/what-is-go...
And you can just sense the hard-earned scars Matt got on the way: at Netflix, Twitter & other places:
Such a good read: newsletter.pragmaticengineer.com/p/what-is-go...
An under-discussed topic: how the hottest software engineering job of the early 2010s is seeing a steady but ongoing decline the last few years.
I'm talking about the native iOS and Android positions. Outside of Big Tech, few startups/scaleups hire for this. Since ~2022?
I'm talking about the native iOS and Android positions. Outside of Big Tech, few startups/scaleups hire for this. Since ~2022?
November 3, 2025 at 3:29 PM
An under-discussed topic: how the hottest software engineering job of the early 2010s is seeing a steady but ongoing decline the last few years.
I'm talking about the native iOS and Android positions. Outside of Big Tech, few startups/scaleups hire for this. Since ~2022?
I'm talking about the native iOS and Android positions. Outside of Big Tech, few startups/scaleups hire for this. Since ~2022?
Another instance of mainstream media reporting on a trend I covered in @pragmaticengineer.com months before.
This time with The Financial Times, 4 months later:
My deepdive in Aug: newsletter.pragmaticengineer.com/p/forward-de...
The FT yesrterday: www.ft.com/content/9100...
This time with The Financial Times, 4 months later:
My deepdive in Aug: newsletter.pragmaticengineer.com/p/forward-de...
The FT yesrterday: www.ft.com/content/9100...
November 3, 2025 at 10:29 AM
Another instance of mainstream media reporting on a trend I covered in @pragmaticengineer.com months before.
This time with The Financial Times, 4 months later:
My deepdive in Aug: newsletter.pragmaticengineer.com/p/forward-de...
The FT yesrterday: www.ft.com/content/9100...
This time with The Financial Times, 4 months later:
My deepdive in Aug: newsletter.pragmaticengineer.com/p/forward-de...
The FT yesrterday: www.ft.com/content/9100...
A dev told me how one of my blog posts helped his book pitch be accepted by a publisher: helping validate broad interest in the topic. The book now getting published!
My blog post was inspired by another post I read online... love how ideas inspire one another in our field
My blog post was inspired by another post I read online... love how ideas inspire one another in our field
November 1, 2025 at 10:21 AM
A dev told me how one of my blog posts helped his book pitch be accepted by a publisher: helping validate broad interest in the topic. The book now getting published!
My blog post was inspired by another post I read online... love how ideas inspire one another in our field
My blog post was inspired by another post I read online... love how ideas inspire one another in our field
OpenAI has started to hire junior software engineers, and it's working great for them.
This comes from me talking with Sulman Choudhry (Head of ChatGPT Engineering). They've started calling them "super juniors" because of the impact they are having newsletter.pragmaticengineer.com/p/san-franci...
This comes from me talking with Sulman Choudhry (Head of ChatGPT Engineering). They've started calling them "super juniors" because of the impact they are having newsletter.pragmaticengineer.com/p/san-franci...
October 31, 2025 at 10:22 AM
OpenAI has started to hire junior software engineers, and it's working great for them.
This comes from me talking with Sulman Choudhry (Head of ChatGPT Engineering). They've started calling them "super juniors" because of the impact they are having newsletter.pragmaticengineer.com/p/san-franci...
This comes from me talking with Sulman Choudhry (Head of ChatGPT Engineering). They've started calling them "super juniors" because of the impact they are having newsletter.pragmaticengineer.com/p/san-franci...
Noticed how a formerly super hip / “break all the rules” GenZ startup with the mission to “make generational wealth” pivoted to being the most boring enterprise SaaS: their website as bland as it gets, “download for Windows”, “Talk to Sales” etc
They realised B2B > B2C for $$
They realised B2B > B2C for $$
October 31, 2025 at 9:01 AM
Noticed how a formerly super hip / “break all the rules” GenZ startup with the mission to “make generational wealth” pivoted to being the most boring enterprise SaaS: their website as bland as it gets, “download for Windows”, “Talk to Sales” etc
They realised B2B > B2C for $$
They realised B2B > B2C for $$
NVIDIA > Amazon + Meta
NVIDIA > Microsoft + Netflix + Palantir
NVIDIA = Google + Meta
This type of cap table math was not on my bingo card
NVIDIA > Microsoft + Netflix + Palantir
NVIDIA = Google + Meta
This type of cap table math was not on my bingo card
October 30, 2025 at 12:48 PM
NVIDIA > Amazon + Meta
NVIDIA > Microsoft + Netflix + Palantir
NVIDIA = Google + Meta
This type of cap table math was not on my bingo card
NVIDIA > Microsoft + Netflix + Palantir
NVIDIA = Google + Meta
This type of cap table math was not on my bingo card
Why does "vibe coding" usually not lead to anything productive? A great person to answer is @addyosmani.bsky.social: working on Chrome for 10+ years, and is the author of the book Beyond Vibe Coding.
Watch or listen:
• YouTube: youtu.be/dHIppEqwi0g
• Spotify: open.spotify.com/episode/12dW...
Watch or listen:
• YouTube: youtu.be/dHIppEqwi0g
• Spotify: open.spotify.com/episode/12dW...
October 29, 2025 at 8:40 PM
Why does "vibe coding" usually not lead to anything productive? A great person to answer is @addyosmani.bsky.social: working on Chrome for 10+ years, and is the author of the book Beyond Vibe Coding.
Watch or listen:
• YouTube: youtu.be/dHIppEqwi0g
• Spotify: open.spotify.com/episode/12dW...
Watch or listen:
• YouTube: youtu.be/dHIppEqwi0g
• Spotify: open.spotify.com/episode/12dW...
AWS becoming the "millennial" cloud? A thing I'm sensing here and there:
Early-stage startups don't bother onboarding to a cloud service like AWS. They instead go for more "hip" startups for web hosting and databases. Think Vercel, Render, Railway, Supabase
They *might* move much later
Early-stage startups don't bother onboarding to a cloud service like AWS. They instead go for more "hip" startups for web hosting and databases. Think Vercel, Render, Railway, Supabase
They *might* move much later
October 29, 2025 at 11:37 AM
AWS becoming the "millennial" cloud? A thing I'm sensing here and there:
Early-stage startups don't bother onboarding to a cloud service like AWS. They instead go for more "hip" startups for web hosting and databases. Think Vercel, Render, Railway, Supabase
They *might* move much later
Early-stage startups don't bother onboarding to a cloud service like AWS. They instead go for more "hip" startups for web hosting and databases. Think Vercel, Render, Railway, Supabase
They *might* move much later
I didn't believe it until I saw it: but it *really* works in open offices!
This is ML engineer Menoua Keshishian at Wispr Flow HQ, in SF, coding... by whispering into his mic (a BOYA Gooseneck that costs ~$70).
I was standing next to him and heard nothing. Everyone in the office does the same
This is ML engineer Menoua Keshishian at Wispr Flow HQ, in SF, coding... by whispering into his mic (a BOYA Gooseneck that costs ~$70).
I was standing next to him and heard nothing. Everyone in the office does the same
October 29, 2025 at 9:08 AM
I didn't believe it until I saw it: but it *really* works in open offices!
This is ML engineer Menoua Keshishian at Wispr Flow HQ, in SF, coding... by whispering into his mic (a BOYA Gooseneck that costs ~$70).
I was standing next to him and heard nothing. Everyone in the office does the same
This is ML engineer Menoua Keshishian at Wispr Flow HQ, in SF, coding... by whispering into his mic (a BOYA Gooseneck that costs ~$70).
I was standing next to him and heard nothing. Everyone in the office does the same
The only two Big Tech companies that neither did mass layoffs when business was booming, nor when things looked bleak:
Apple** and NVIDIA
All others (Amazon, Google, Meta) did (and do) mass cuts as business is booming. Corporate reality
**Apple not since 1997 that is
Apple** and NVIDIA
All others (Amazon, Google, Meta) did (and do) mass cuts as business is booming. Corporate reality
**Apple not since 1997 that is
October 28, 2025 at 12:33 PM
The only two Big Tech companies that neither did mass layoffs when business was booming, nor when things looked bleak:
Apple** and NVIDIA
All others (Amazon, Google, Meta) did (and do) mass cuts as business is booming. Corporate reality
**Apple not since 1997 that is
Apple** and NVIDIA
All others (Amazon, Google, Meta) did (and do) mass cuts as business is booming. Corporate reality
**Apple not since 1997 that is
I keep being amused by some devs claiming that multi-region support for a service should be easy to add. Then turns out they're unaware of what active-active arch is.
Multi-region support is not all that complex from afar, but it gets complicated with the details.
Multi-region support is not all that complex from afar, but it gets complicated with the details.
October 28, 2025 at 8:03 AM
I keep being amused by some devs claiming that multi-region support for a service should be easy to add. Then turns out they're unaware of what active-active arch is.
Multi-region support is not all that complex from afar, but it gets complicated with the details.
Multi-region support is not all that complex from afar, but it gets complicated with the details.
One realization from the AWS outage:
Those saying Vercel is "just" an AWS wrapper that charges more than AWS were... wrong.
Vercel failed over and recovered ~10 hours before the AWS outage was over. Them (and other IaaS / PaaS platforms) invest into multi-region resillience...
Those saying Vercel is "just" an AWS wrapper that charges more than AWS were... wrong.
Vercel failed over and recovered ~10 hours before the AWS outage was over. Them (and other IaaS / PaaS platforms) invest into multi-region resillience...
October 26, 2025 at 10:16 PM
One realization from the AWS outage:
Those saying Vercel is "just" an AWS wrapper that charges more than AWS were... wrong.
Vercel failed over and recovered ~10 hours before the AWS outage was over. Them (and other IaaS / PaaS platforms) invest into multi-region resillience...
Those saying Vercel is "just" an AWS wrapper that charges more than AWS were... wrong.
Vercel failed over and recovered ~10 hours before the AWS outage was over. Them (and other IaaS / PaaS platforms) invest into multi-region resillience...
I hate how airlines treat their software department as a cost center - and as a result, their websites are unusuable for edge cases; but there is no monitoring on these; passengers suffer, devs don't know or care.
This is Delta, when trying to submit a reimbursement request 👎
This is Delta, when trying to submit a reimbursement request 👎
October 25, 2025 at 2:59 PM
I hate how airlines treat their software department as a cost center - and as a result, their websites are unusuable for edge cases; but there is no monitoring on these; passengers suffer, devs don't know or care.
This is Delta, when trying to submit a reimbursement request 👎
This is Delta, when trying to submit a reimbursement request 👎
It's great that AWS already shared a postmortem, and we can understand some things that went wrong, starting the outage.
A summary of the events, based on what the AWS team shared:
A summary of the events, based on what the AWS team shared:
What caused the AWS outage that felt like it took down half the internet on Monday?
It started with a race condition in DynamoDB's DNS propagation, which ended up setting the dynamod.us-east-1.amazonaws.com DNS to empty.
A deepdive and more details: blog.pragmaticengineer.com/aws-outage-u...
It started with a race condition in DynamoDB's DNS propagation, which ended up setting the dynamod.us-east-1.amazonaws.com DNS to empty.
A deepdive and more details: blog.pragmaticengineer.com/aws-outage-u...
October 24, 2025 at 2:45 PM
It's great that AWS already shared a postmortem, and we can understand some things that went wrong, starting the outage.
A summary of the events, based on what the AWS team shared:
A summary of the events, based on what the AWS team shared:
Can we for a second appreciate how Robinhood took no responsibility for being down, and tried to put the blame on AWS.
However, it was Robinhood that decided to run from a single AWS region, and not be preapred to fail over, like other, more resillient AWS users did...
However, it was Robinhood that decided to run from a single AWS region, and not be preapred to fail over, like other, more resillient AWS users did...
October 23, 2025 at 5:17 PM
Can we for a second appreciate how Robinhood took no responsibility for being down, and tried to put the blame on AWS.
However, it was Robinhood that decided to run from a single AWS region, and not be preapred to fail over, like other, more resillient AWS users did...
However, it was Robinhood that decided to run from a single AWS region, and not be preapred to fail over, like other, more resillient AWS users did...
The Wall Street Journal is coveing today the same trend I shared in @pragmaticengineer.com, back in August: extreme hours at AI startups.covering
Pretty validating when mainstream media covers the same trends months later!
(And a reason to be subscribed to The Pragmatic Engineer!)
Pretty validating when mainstream media covers the same trends months later!
(And a reason to be subscribed to The Pragmatic Engineer!)
October 23, 2025 at 3:11 PM
The Wall Street Journal is coveing today the same trend I shared in @pragmaticengineer.com, back in August: extreme hours at AI startups.covering
Pretty validating when mainstream media covers the same trends months later!
(And a reason to be subscribed to The Pragmatic Engineer!)
Pretty validating when mainstream media covers the same trends months later!
(And a reason to be subscribed to The Pragmatic Engineer!)
No way the UK's tax authority (HMRC) was running their operations off of a US data center
... but they were!
AWS us-east-1 outage knocked them offline, and exposed relying on US infra to serve the British public
Ironic from a government with ample home grown DCs + talent
... but they were!
AWS us-east-1 outage knocked them offline, and exposed relying on US infra to serve the British public
Ironic from a government with ample home grown DCs + talent
October 23, 2025 at 10:30 AM
No way the UK's tax authority (HMRC) was running their operations off of a US data center
... but they were!
AWS us-east-1 outage knocked them offline, and exposed relying on US infra to serve the British public
Ironic from a government with ample home grown DCs + talent
... but they were!
AWS us-east-1 outage knocked them offline, and exposed relying on US infra to serve the British public
Ironic from a government with ample home grown DCs + talent