I know the fourth rail on HN these days besides sex, religion and politics seems to be “AI”, but AI has taken the drudgery out of going from design -> implementation for me at least.
this is a very interesting line of analogy .. allow me to broaden it with "a backhoe cannot do the mantle inlays with black oak" and at the same time "your backhoe is done in four days but the entire home still needs building"
These metaphors are always so facile - if a backhoe dug a randomized foundation for your house (all the while apologizing and re-digging) I'm pretty sure you would not use it over the shovel. A better analogy would've been using dynamite to dig a foundation over using a shovel.
Every project I’ve done that integrates with and LLM and/or used for coding the customer has paid my employer for and I have never gotten a less than perfect CSAT score from my customer or my PM. They have all been done on time, on budget and met both the functional and non functional requirements.
I’ve implemented LLM based “intent processing” instead of old school “give it a list of every possible phrase that someone could use to activate an intent” for call centers for a couple of states and a couple colleges - I worked for AWS ProServe directly in the pub sec division and now work for a third party consulting company.
Think of an intent as something like asking Siri for driving directions, setting an alarm etc.
One of my specialties is Amazon Connect for hosted call centers - based on the same service that Amazon uses internally. Think of the difference between how Google Assistant and Alexa worked pre LLM and how Siri works today compared to tool calling that modern LLMs used.
That’s just using an LLM in a product.
As far as how I use it everyday? I am a staff consultant and in the previous times, I needed at least a couple of developers under me just to get the grunt work done on time. Now I can just treat Claude Code and Codex as faster, more accurate ticker takers.
Of course I’m not going to dox myself (more than I probably already have) or break any NDAs
And before the gate keeping starts about “I must be inexperienced”. My first time coding was on a 1Mhz Apple //e in assembly and BASIC in 1986. During the first decade and a half of my professional career, I programmed in C targeting everything from mainframes to PCs to Windows CE ruggedized devices.
Curious if today's Berkeley's professors would still wear Alphabet (former Google) t-shirts while holding presentations, I now realise that things have changed a lot in the last 10 years.
I've also not gone through the whole presentation, but does he at any point talk about the moral choices one will most definitely have to make during a career in tech? (this is related to the previous paragraph). Is it a "bad career" if people choose not to work for companies (such as Alphabet) that have gone all in behind AI? Seeing as now AI is used by State-entities for very nefarious reasons. Like I said, 2026 is way different compared to 2016.
I’ve been thinking a lot more about this lately. Big Tech today is far more powerful than 1990s Microsoft and 1970s IBM ever were. I’m not anti-AI, but the sheer power major players like OpenAI, Alphabet, Microsoft, and Meta have make me very nervous.
The challenge for computer science researchers who have qualms about working for Big Tech is finding an alternative career path. Speaking from an American point of view, academia has always been competitive, and the immediate future of research funding is uncertain given the political climate. This uncertainty also extends to government labs. The challenge with industry research is that there are not a lot of non-Big Tech employers of computer science researchers. This leaves starting a business, but business is very different from research.
I’m a tenure-track professor at a community college in the Bay Area. While I’ll never be able to afford to purchase a home near my job, I am able to live well as a single man renting an apartment. I have a great career teaching and using my long summer breaks for research and side projects. I like not having to worry about “publish or perish,” and I enjoy teaching and mentoring students. While this might not be considered “successful” for some people who are aiming for a professorship at an R1 university or an industry job at a top company’s top lab, I love my job and believe it’s a fantastic route for someone who enjoys teaching and who also wants extended time during the summer for research and side projects.
Big Tech today is far more powerful than 1990s Microsoft and 1970s IBM ever were.
In aggregate, sure, but no company today comes within an order of magnitude of the power an IBM of the ‘70s and ‘80s or a Microsoft of the ‘90s and ‘00s had over the tech landscape.
1970s IBM and 1990s Microsoft were formidable monopolies, but I was thinking in lines of influence over society and not necessarily in terms of market share. The consequences of social media and centralized Web services are much more impactful on society, for better and for worse, than dominance over 1970s mainframes and 1990s desktop operating systems, Web browsers, and office suites.
That power isn't even a rounding error compared to their power today. In those decades you had commercial dominance over a niche sector. Today you play kingmaker together with the other oligarchs.
Big Tech today is the media. Taken together, they completely control what a majority of the populace knows about the world. It is considered completely impractical and somewhat suspicious not to carry one of their location tracking communication devices at all times.
Orwell's Ministry of Truth could not dream of what Meta, Alphabet, OpenAI and Apple can do at any time, anywhere.
Maybe it's my own bias (after working for a big tech company for 5 years), but it's certainly not the dream job I thought it would be. Soulless corporations with questionable impact on society, lot of turnover, increasing pressure every year, fear of layoffs. Even the tech isn't that exciting, there's lots of tedious work, technical debt, hacked solutions, no time for researching and building quality solution.
That being said, it's certainly different for researchers. I can imagine that being a researcher at Google is more fun than being a median SWE in another FAANG. But still, I find these companies less enticing in general, even the products tend to degrade as they keep pushing the monetization.
I’m an ex-industry researcher with experience at FAANGs, albeit as a software engineering intern (Google) and a production engineer (Facebook).
I think it depends on the interests of the researcher. If a researcher is comfortable being a “brain for hire” who is comfortable solving research problems that are driven by business needs and where there needs to be short-term or medium-term results, then I think there are plenty of opportunities at large companies, including the FAANGs. I find research more fun than software engineering, but researchers are far from immune from pressures to ship.
If a researcher is more interested in curiosity-driven work and who wants to work on a longer time frame, I’m afraid that there’s no place in industry, except for maybe Microsoft Research (which I’ve heard changed under Satya Nadella), that supports such work. The days of Bob Taylor-era Xerox PARC and Unix-era Bell Labs ended many decades ago, and while there were still curiosity-driven labs in industry well into the 2010s, I have witnessed the remainder of these old-style labs change their missions to become much more focused on immediate and near-immediate business needs.
My experience as an L5 Google Research research scientist is that I have a lot of freedom as long as I can show I've made progress on one or two things the company cares about at each annual review. IMO, this is about as much research freedom as is reasonable for me to request.
15 - 20 years ago i was a total Google fanboy, i liked what they did with their search, i loved their overall ecosystem and perceived culture. I even have a - now orphaned - googlemail email address.
Nowadays? I wouldn't touch anything that comes from Google (granted also not from any other big tech company) with pliers.
As he's mentioned, he has given this talk a few times, aimed at different audiences. One version of the talk has its slides here: https://drive.google.com/file/d/0Bzis5MXW83vCdUdXYnFIVDVOSkE...
> Successful projects go through an unglamorous hard phase.
> Design is more fun than making it work.
Great wisdom for any kind of project
hello, yes 911? I'd like to report a hate crime against me
I know the fourth rail on HN these days besides sex, religion and politics seems to be “AI”, but AI has taken the drudgery out of going from design -> implementation for me at least.
I think the issue is that some of us like the drudgery!
Some of us like lifting weights too but you can't fault someone for using a backhoe instead of a shovel when they want to build a house.
this is a very interesting line of analogy .. allow me to broaden it with "a backhoe cannot do the mantle inlays with black oak" and at the same time "your backhoe is done in four days but the entire home still needs building"
A backhoe cannot do the mantle inlays with black oak, but maybe a CNC router can do most of the work for you except the final detailing and fitting.
Will I still like it?
Once pricing comes into play, you might.
These metaphors are always so facile - if a backhoe dug a randomized foundation for your house (all the while apologizing and re-digging) I'm pretty sure you would not use it over the shovel. A better analogy would've been using dynamite to dig a foundation over using a shovel.
More like house flippers drywalling over the bad wiring and weird framing. "Looks good!"
Unfortunately, even if I did like the “drudgery” in consulting, no one pays American rates for it. The drudgery is done by folks in LatAm and India.
World is bigger than US, much bigger. So is HN audience.
And in the rest of the world do the people doing the “drudgery” get paid more than the people who lead the work and design?
Great! Can you point us to a public successful example of your work using AI for implementation?
Every project I’ve done that integrates with and LLM and/or used for coding the customer has paid my employer for and I have never gotten a less than perfect CSAT score from my customer or my PM. They have all been done on time, on budget and met both the functional and non functional requirements.
I’ve implemented LLM based “intent processing” instead of old school “give it a list of every possible phrase that someone could use to activate an intent” for call centers for a couple of states and a couple colleges - I worked for AWS ProServe directly in the pub sec division and now work for a third party consulting company.
Think of an intent as something like asking Siri for driving directions, setting an alarm etc.
One of my specialties is Amazon Connect for hosted call centers - based on the same service that Amazon uses internally. Think of the difference between how Google Assistant and Alexa worked pre LLM and how Siri works today compared to tool calling that modern LLMs used.
That’s just using an LLM in a product.
As far as how I use it everyday? I am a staff consultant and in the previous times, I needed at least a couple of developers under me just to get the grunt work done on time. Now I can just treat Claude Code and Codex as faster, more accurate ticker takers.
Of course I’m not going to dox myself (more than I probably already have) or break any NDAs
And before the gate keeping starts about “I must be inexperienced”. My first time coding was on a 1Mhz Apple //e in assembly and BASIC in 1986. During the first decade and a half of my professional career, I programmed in C targeting everything from mainframes to PCs to Windows CE ruggedized devices.
It's always crickets
It’s called NDAs and I in fact did reply. Hours before yours
thank you for posting a text version. to make the topic even more legible to prospective link clickers:
> How to Have a Bad Career in Research/Academia Pre-PhD and Post-PhD (& How to Give a Bad Talk)
Curious if today's Berkeley's professors would still wear Alphabet (former Google) t-shirts while holding presentations, I now realise that things have changed a lot in the last 10 years.
I've also not gone through the whole presentation, but does he at any point talk about the moral choices one will most definitely have to make during a career in tech? (this is related to the previous paragraph). Is it a "bad career" if people choose not to work for companies (such as Alphabet) that have gone all in behind AI? Seeing as now AI is used by State-entities for very nefarious reasons. Like I said, 2026 is way different compared to 2016.
I’ve been thinking a lot more about this lately. Big Tech today is far more powerful than 1990s Microsoft and 1970s IBM ever were. I’m not anti-AI, but the sheer power major players like OpenAI, Alphabet, Microsoft, and Meta have make me very nervous.
The challenge for computer science researchers who have qualms about working for Big Tech is finding an alternative career path. Speaking from an American point of view, academia has always been competitive, and the immediate future of research funding is uncertain given the political climate. This uncertainty also extends to government labs. The challenge with industry research is that there are not a lot of non-Big Tech employers of computer science researchers. This leaves starting a business, but business is very different from research.
I’m a tenure-track professor at a community college in the Bay Area. While I’ll never be able to afford to purchase a home near my job, I am able to live well as a single man renting an apartment. I have a great career teaching and using my long summer breaks for research and side projects. I like not having to worry about “publish or perish,” and I enjoy teaching and mentoring students. While this might not be considered “successful” for some people who are aiming for a professorship at an R1 university or an industry job at a top company’s top lab, I love my job and believe it’s a fantastic route for someone who enjoys teaching and who also wants extended time during the summer for research and side projects.
Big Tech today is far more powerful than 1990s Microsoft and 1970s IBM ever were.
In aggregate, sure, but no company today comes within an order of magnitude of the power an IBM of the ‘70s and ‘80s or a Microsoft of the ‘90s and ‘00s had over the tech landscape.
1970s IBM and 1990s Microsoft were formidable monopolies, but I was thinking in lines of influence over society and not necessarily in terms of market share. The consequences of social media and centralized Web services are much more impactful on society, for better and for worse, than dominance over 1970s mainframes and 1990s desktop operating systems, Web browsers, and office suites.
Probably true but what was tech is now basically just the entire economy.
That power isn't even a rounding error compared to their power today. In those decades you had commercial dominance over a niche sector. Today you play kingmaker together with the other oligarchs.
Big Tech today is the media. Taken together, they completely control what a majority of the populace knows about the world. It is considered completely impractical and somewhat suspicious not to carry one of their location tracking communication devices at all times.
Orwell's Ministry of Truth could not dream of what Meta, Alphabet, OpenAI and Apple can do at any time, anywhere.
That's a good observation. It's certainly less glamorous to work for these companies nowadays, but it started before AI.
Is it less glamorous? As an academic, my impression is that lots of innovation is now done at industrial labs.
In general, Academia lacks sufficient resources and appropriate structures for dedicated efforts.
Pay is abysmal and politics is toxic. Both scare away lots of technical talent.
Maybe it's my own bias (after working for a big tech company for 5 years), but it's certainly not the dream job I thought it would be. Soulless corporations with questionable impact on society, lot of turnover, increasing pressure every year, fear of layoffs. Even the tech isn't that exciting, there's lots of tedious work, technical debt, hacked solutions, no time for researching and building quality solution.
That being said, it's certainly different for researchers. I can imagine that being a researcher at Google is more fun than being a median SWE in another FAANG. But still, I find these companies less enticing in general, even the products tend to degrade as they keep pushing the monetization.
I’m an ex-industry researcher with experience at FAANGs, albeit as a software engineering intern (Google) and a production engineer (Facebook).
I think it depends on the interests of the researcher. If a researcher is comfortable being a “brain for hire” who is comfortable solving research problems that are driven by business needs and where there needs to be short-term or medium-term results, then I think there are plenty of opportunities at large companies, including the FAANGs. I find research more fun than software engineering, but researchers are far from immune from pressures to ship.
If a researcher is more interested in curiosity-driven work and who wants to work on a longer time frame, I’m afraid that there’s no place in industry, except for maybe Microsoft Research (which I’ve heard changed under Satya Nadella), that supports such work. The days of Bob Taylor-era Xerox PARC and Unix-era Bell Labs ended many decades ago, and while there were still curiosity-driven labs in industry well into the 2010s, I have witnessed the remainder of these old-style labs change their missions to become much more focused on immediate and near-immediate business needs.
My experience as an L5 Google Research research scientist is that I have a lot of freedom as long as I can show I've made progress on one or two things the company cares about at each annual review. IMO, this is about as much research freedom as is reasonable for me to request.
If this was 2016, wasn't he employed by Google at the time?
One of them (Patterson & Hennessy) was, and the other was (is?) chairman of the board, I forget which.
15 - 20 years ago i was a total Google fanboy, i liked what they did with their search, i loved their overall ecosystem and perceived culture. I even have a - now orphaned - googlemail email address.
Nowadays? I wouldn't touch anything that comes from Google (granted also not from any other big tech company) with pliers.
What do you use for search?
https://metager.de https://wiby.me
(in Academia)