Jake Gold
@jacob.gold
Infra eng leader, SF Bay Area. Helped launch and scale Bluesky. Prev: Nuro, Docker, Google, founder.
Currently working on AI codegen.
Obsessed with history, computers, and open systems.
Happy to chat: bluesky@jacob.gold
Currently working on AI codegen.
Obsessed with history, computers, and open systems.
Happy to chat: bluesky@jacob.gold
A man could hack in a place like that.
November 9, 2025 at 5:34 AM
A man could hack in a place like that.
If you haven't seen it, a fun webapp that is entirely free supernative.tv/ja/
It plays Japanese TV drama clips and asks you to type in words that you hear. I'm super rusty these days but I found it very helpful.
It plays Japanese TV drama clips and asks you to type in words that you hear. I'm super rusty these days but I found it very helpful.
November 9, 2025 at 12:00 AM
If you haven't seen it, a fun webapp that is entirely free supernative.tv/ja/
It plays Japanese TV drama clips and asks you to type in words that you hear. I'm super rusty these days but I found it very helpful.
It plays Japanese TV drama clips and asks you to type in words that you hear. I'm super rusty these days but I found it very helpful.
Haha, agreed, I’ve compared it to financial advisors and bitcoin miners in the same way.
Truth is that the tech is very powerful but only as a tool for humans for now.
So it makes sense to sell because the alternative is keeping the tech and hiring many millions of people, which is impractical.
Truth is that the tech is very powerful but only as a tool for humans for now.
So it makes sense to sell because the alternative is keeping the tech and hiring many millions of people, which is impractical.
November 8, 2025 at 2:43 PM
Haha, agreed, I’ve compared it to financial advisors and bitcoin miners in the same way.
Truth is that the tech is very powerful but only as a tool for humans for now.
So it makes sense to sell because the alternative is keeping the tech and hiring many millions of people, which is impractical.
Truth is that the tech is very powerful but only as a tool for humans for now.
So it makes sense to sell because the alternative is keeping the tech and hiring many millions of people, which is impractical.
20+ which is reduced from the peak of my illness.
November 8, 2025 at 2:08 PM
20+ which is reduced from the peak of my illness.
Yeah, exactly.
I loved being able to wear a mask in Japan pre-Covid but didn’t want to freak people out elsewhere by wearing one.
I loved being able to wear a mask in Japan pre-Covid but didn’t want to freak people out elsewhere by wearing one.
November 8, 2025 at 2:06 PM
Yeah, exactly.
I loved being able to wear a mask in Japan pre-Covid but didn’t want to freak people out elsewhere by wearing one.
I loved being able to wear a mask in Japan pre-Covid but didn’t want to freak people out elsewhere by wearing one.
Theoretically but not if a sick person is close enough to you or the filtration isn’t running during boarding or deplaning etc.
November 8, 2025 at 2:01 PM
Theoretically but not if a sick person is close enough to you or the filtration isn’t running during boarding or deplaning etc.
What do you mean by profitable here? It's not clear.
One can run gpt-oss-120b on a Macbook and do highly productive paid programming work with it. The cost is effectively zero.
These LLM inference programs are just computer programs. Of course we'll be able to run them faster/cheaper each year.
One can run gpt-oss-120b on a Macbook and do highly productive paid programming work with it. The cost is effectively zero.
These LLM inference programs are just computer programs. Of course we'll be able to run them faster/cheaper each year.
November 8, 2025 at 6:29 AM
What do you mean by profitable here? It's not clear.
One can run gpt-oss-120b on a Macbook and do highly productive paid programming work with it. The cost is effectively zero.
These LLM inference programs are just computer programs. Of course we'll be able to run them faster/cheaper each year.
One can run gpt-oss-120b on a Macbook and do highly productive paid programming work with it. The cost is effectively zero.
These LLM inference programs are just computer programs. Of course we'll be able to run them faster/cheaper each year.
Yeah, I don't think even Altman thinks LLMs are the thing but he believes (prays to god) they're enough to get us to the next step, which will get us to the next step, etc.
Seems more plausible that some clever person funded by all this AI investment money will make a new surprising breakthrough.
Seems more plausible that some clever person funded by all this AI investment money will make a new surprising breakthrough.
November 8, 2025 at 4:51 AM
Yeah, I don't think even Altman thinks LLMs are the thing but he believes (prays to god) they're enough to get us to the next step, which will get us to the next step, etc.
Seems more plausible that some clever person funded by all this AI investment money will make a new surprising breakthrough.
Seems more plausible that some clever person funded by all this AI investment money will make a new surprising breakthrough.
Yeah translation is so cheap now that it's not even considered worth charging for. That's where everything we have today is headed.
The Big Question is whether anyone is clever enough to do something far beyond LLMs.
Sam Altman's concept that that LLMs will help us is create nextgen AI is dubious.
The Big Question is whether anyone is clever enough to do something far beyond LLMs.
Sam Altman's concept that that LLMs will help us is create nextgen AI is dubious.
November 8, 2025 at 4:20 AM
Yeah translation is so cheap now that it's not even considered worth charging for. That's where everything we have today is headed.
The Big Question is whether anyone is clever enough to do something far beyond LLMs.
Sam Altman's concept that that LLMs will help us is create nextgen AI is dubious.
The Big Question is whether anyone is clever enough to do something far beyond LLMs.
Sam Altman's concept that that LLMs will help us is create nextgen AI is dubious.
There's obviously a chance I've had Covid and been asymptomatic, but yeah it seems be pretty effective, and doesn't cause me any real inconvenience.
Dan didn't realize he was sick until later in the night and then took a test and we all had a good laugh about it.
Dan didn't realize he was sick until later in the night and then took a test and we all had a good laugh about it.
November 8, 2025 at 4:16 AM
There's obviously a chance I've had Covid and been asymptomatic, but yeah it seems be pretty effective, and doesn't cause me any real inconvenience.
Dan didn't realize he was sick until later in the night and then took a test and we all had a good laugh about it.
Dan didn't realize he was sick until later in the night and then took a test and we all had a good laugh about it.
If OpenAI only offered smaller models on ChatGPT, rather than their new/huge ones, it could already be one of the most profitable products in the world.
They can't do this because users would just switch to the competition's bigger models.
Over time, new hardware will make current models cheap.
They can't do this because users would just switch to the competition's bigger models.
Over time, new hardware will make current models cheap.
November 8, 2025 at 4:14 AM
If OpenAI only offered smaller models on ChatGPT, rather than their new/huge ones, it could already be one of the most profitable products in the world.
They can't do this because users would just switch to the competition's bigger models.
Over time, new hardware will make current models cheap.
They can't do this because users would just switch to the competition's bigger models.
Over time, new hardware will make current models cheap.
(Doc/hospital most of the time too)
It's really the airport/jetway/queuing areas that seem the highest risk when flying, as much as the plane itself.
I'm not terrified of getting Covid but when I'm traveling for fun or work I really don't want to be out of commission for days in a hotel room.
It's really the airport/jetway/queuing areas that seem the highest risk when flying, as much as the plane itself.
I'm not terrified of getting Covid but when I'm traveling for fun or work I really don't want to be out of commission for days in a hotel room.
November 8, 2025 at 4:09 AM
(Doc/hospital most of the time too)
It's really the airport/jetway/queuing areas that seem the highest risk when flying, as much as the plane itself.
I'm not terrified of getting Covid but when I'm traveling for fun or work I really don't want to be out of commission for days in a hotel room.
It's really the airport/jetway/queuing areas that seem the highest risk when flying, as much as the plane itself.
I'm not terrified of getting Covid but when I'm traveling for fun or work I really don't want to be out of commission for days in a hotel room.
Yeah I wouldn't want to be an undifferentiated AI model company. I think the ChatGPT app/web site is the most valuable asset at OpenAI, not their models, teams, or secret projects.
We'll definitely get 2025 level models on local soon enough. Big question is whether cloud models will be much better.
We'll definitely get 2025 level models on local soon enough. Big question is whether cloud models will be much better.
November 8, 2025 at 3:57 AM
Yeah I wouldn't want to be an undifferentiated AI model company. I think the ChatGPT app/web site is the most valuable asset at OpenAI, not their models, teams, or secret projects.
We'll definitely get 2025 level models on local soon enough. Big question is whether cloud models will be much better.
We'll definitely get 2025 level models on local soon enough. Big question is whether cloud models will be much better.
The bet that AI won't be profitable is almost guaranteed to be wrong.
Current models are already at the "magic" level and it won't take many years before we have AI accelerator chips that are orders of magnitude more efficient than current ones.
That's where the big bet confidence comes from IMHO.
Current models are already at the "magic" level and it won't take many years before we have AI accelerator chips that are orders of magnitude more efficient than current ones.
That's where the big bet confidence comes from IMHO.
November 8, 2025 at 3:37 AM
The bet that AI won't be profitable is almost guaranteed to be wrong.
Current models are already at the "magic" level and it won't take many years before we have AI accelerator chips that are orders of magnitude more efficient than current ones.
That's where the big bet confidence comes from IMHO.
Current models are already at the "magic" level and it won't take many years before we have AI accelerator chips that are orders of magnitude more efficient than current ones.
That's where the big bet confidence comes from IMHO.
Pretty sure I registered my first one in 1996 as a teenager. I did let that one expire and I regret it!
November 8, 2025 at 3:33 AM
Pretty sure I registered my first one in 1996 as a teenager. I did let that one expire and I regret it!
The best tid to plant a merkle search tree is 20 years ago, the second best tid is today.
November 8, 2025 at 3:31 AM
The best tid to plant a merkle search tree is 20 years ago, the second best tid is today.
Just today I looked at my Namecheap account and realized I've got 20 year old domains I've been paying for that I don't need for anything and probably never will.
Did I cancel them? Of course not.
Did I cancel them? Of course not.
November 8, 2025 at 3:29 AM
Just today I looked at my Namecheap account and realized I've got 20 year old domains I've been paying for that I don't need for anything and probably never will.
Did I cancel them? Of course not.
Did I cancel them? Of course not.
These companies already have access to government-level money so can't imagine it would be necessary in any circumstance.
It might even be painful-but-healthy to have a crash in the AI market to keep things from getting overheated.
In retrospect the dot com crash was probably a net-positive thing.
It might even be painful-but-healthy to have a crash in the AI market to keep things from getting overheated.
In retrospect the dot com crash was probably a net-positive thing.
November 8, 2025 at 3:27 AM
These companies already have access to government-level money so can't imagine it would be necessary in any circumstance.
It might even be painful-but-healthy to have a crash in the AI market to keep things from getting overheated.
In retrospect the dot com crash was probably a net-positive thing.
It might even be painful-but-healthy to have a crash in the AI market to keep things from getting overheated.
In retrospect the dot com crash was probably a net-positive thing.
Investors have plenty of money to buy up whatever is valuable if an AI model company fails.
So there's no need for the government to step in.
And at least so far, none of the AI model companies has any uniquely special technology. The technology is currently a commodity.
So there's no need for the government to step in.
And at least so far, none of the AI model companies has any uniquely special technology. The technology is currently a commodity.
November 8, 2025 at 3:15 AM
Investors have plenty of money to buy up whatever is valuable if an AI model company fails.
So there's no need for the government to step in.
And at least so far, none of the AI model companies has any uniquely special technology. The technology is currently a commodity.
So there's no need for the government to step in.
And at least so far, none of the AI model companies has any uniquely special technology. The technology is currently a commodity.
In fairness, women did create these men, so...
November 7, 2025 at 12:31 AM
In fairness, women did create these men, so...
Well he might have an owner end up back home? I don't want to keep anyone's lost pet that they miss.
November 7, 2025 at 12:05 AM
Well he might have an owner end up back home? I don't want to keep anyone's lost pet that they miss.