4 ms·

So he's spent $51k on tokens in 3 months to build what exactly? Tools to enable you to spend more money on tokens? Quick math on the environmental impact of th

by ossa-ma 9mo ago

So he's spent $51k on tokens in 3 months to build what exactly? Tools to enable you to spend more money on tokens? Quick math on the environmental impact of this assuming 18.35Wh/1000 tokens: Total energy: 4.73GWh, equivalent of powering 450 average US homes annually Carbon footprint: ~1822 metric tons of CO2, equivalent of driving 4.56 million miles in a gas powered car Water consumption: 4.5 million litres, recommended daily water intake for 4000 people for a full year Yet they're on twitter bragging... https://x.com/steipete/status/2004675874499842535 https://x.com/steipete/status/2004675874499842535

wg0 9mo ago
Well. TBH for some maybe this is the wrong question to ask but I have been thinking where did those 250 billion tokens go? What tools/products/services came out of that? EDIT: Typos
- rocmcd 9mo ago
  This has been my biggest question through this whole AI craze. If AI is making everyone X% more productive, where is the proof? Shouldn't we expect to see new startups now able to compete with large enterprises? Shouldn't we be seeing new amazing apps? New features being added? Shouldn't bugs be a thing of the past? Shouldn't uptime be a solved problem by now? I look around and everything seems to be... the same? Apart from the availability of these AI tools, what has meaningfully changed since 2020?
  - Sevii 9mo ago
    AI coding improved a lot over 2025. In early 2025 LLMs still struggled with counting. Now they are capable of tool calling so they can just use a calculator. Frankly, I'd say AI coding may as well have not existed before mid-2025. The output wasn't really that good. Sure you could generate code but couldn't rely on a coding agent to make 2 line edits to a 1000 line file.
    - rocmcd 9mo ago
      I don't doubt that they have improved a lot this year, but the same claims were being made last year as well. And the year before that. I still haven't seen anything that proves to me that people are truly that much more productive. They certainly _feel_ more productive, though. Hell, the GP spent more than $50,000 this year on API calls alone and the results are... what again? Where is the innovation? Where are the tools that wouldn't have been possible to build pre-ChatGPT? I'm constantly reminded of the Feynman quote: "The first principle is that you must not fool yourself, and you are the easiest person to fool."
croemer 9mo ago
What's the source for the energy per token? I guess this? https://www.theguardian.com/technology/2025/aug/09/open-ai-chat-gpt5-energy-use?utm_source=chatgpt.com https://www.theguardian.com/technology/2025/aug/09/open-ai-c... 18Wh/1000t is on the high end of estimates. But even if it's 10x less I agree this is pretty crazy usage.
- ossa-ma 9mo ago
  "The University of Rhode Island based its report on its estimates that producing a medium-length, 1,000-token GPT-5 response can consume up to 40 watt-hours (Wh) of electricity, with an average just over 18.35 Wh, up from 2.12 Wh for GPT-4. This was higher than all other tested models, except for OpenAI's o3 (25.35 Wh) and Deepseek's R1 (20.90 Wh)." https://www.tomshardware.com/tech-industry/artificial-intelligence/chatgpt-5-power-consumption-could-be-as-much-as-eight-times-higher-than-gpt-4-research-institute-estimates-medium-sized-gpt-5-response-can-consume-up-to-40-watt-hours-of-electricity https://www.tomshardware.com/tech-industry/artificial-intell... https://app.powerbi.com/view?r=eyJrIjoiZjVmOTI0MmMtY2U2Mi00ZTE2LTk2MGYtY2ZjNDMzODZkMjlmIiwidCI6IjQyNmQyYThkLTljY2QtNDI1NS04OTNkLTA2ODZhMzJjMTY4ZCIsImMiOjF9 https://app.powerbi.com/view?r=eyJrIjoiZjVmOTI0MmMtY2U2Mi00Z... https://blog.samaltman.com/the-gentle-singularity https://blog.samaltman.com/the-gentle-singularity
  - causal 9mo ago
    These numbers don't pass sanity check for me. With 4x300W cards you can get a 1K token DeepSeek R1 output in about 10 seconds. That's just 3.3Wh right? And that's before you even consider batching.
azan_ 9mo ago
That's very dishonest. Daily water intake is a fraction of how people use water! Producing 1 kg of beef requires 15 000 litres and if you put it that way (which is much more honest) it's not that bad. If you'd also take into account other ways people use water then it'd be even less shocking.
- ossa-ma 9mo ago
  And the other stats you ignored? As well as the main point I was making?
  - azan_ 9mo ago
    I don't know about other stats, that's why I won't comment on them. But it doesn't matter - your water use stats are still manipulative and make your point much weaker.
    - ossa-ma 9mo ago
      I've adjusted the statistic for you: Adjusted beef consumption: 4.5 million litres of water can be used to produce 300kg of beef -> US (highest beef consumer/capita) consumes 23.3kg of beef , enough to feed ~13 Americans (30 Brits, ~43 Japanese) yummy delicious grass-fed beef yearly!
      - elevation 9mo ago
        These number seem off. A single head of cattle may contain 300kg of edible beef with a hanging weight of roughly twice that. In what world does raising a single bovine consume of 4.5 million liters of water? Neither the cow nor the cow's food retains much water; the water is merely delayed a little in its journey to the local watershed, and in vast parts of the US, local rainfall is adequate for this purpose (power irrigation isn't required for the crops, and cattle may drink from a pond.) Even if a cow drinks pumped well water, the majority of its nourishment will have been itself sustained by local natural rainfall. A datacenter's use of water over any timescale can hardly be compared with a cow's.
gwern 9mo ago
> So he's spent $51k on tokens in 3 months to build what exactly? Tools to enable you to spend more money on tokens? Sounds like that's more than a junior or an intern that would have cost twice as much in fully loaded cost.
- aprilthird2021 9mo ago
  He also put his own time into it. It's also more than hiring someone overseas esp just for a few months. Honestly it's more than most interns are paid for 3 months outside FAANG (considering housing is paid for there etc)
  - gwern 9mo ago
    1. You put a lot of time into an intern or a junior too. 2. I didn't say 'paid', I said, 'fully loaded [total] cost'. The total cost of them goes far beyond their mere salary - the search process like all of the interviews for all candidates, onboarding, HR, taxes etc.
    - aprilthird2021 9mo ago
      1. Idk, I didn't have to. I managed an intern a few months back and he just did everything we had planned out and written down then started making his own additions on top. 2. Yeah I mentioned that also. 3. It's still more expensive than hiring a contractor esp abroad, even all in.
jeswin 9mo ago
I spent ~$1500 (and a bit more than two months) building a compiler that translates typescript to native binaries. I think I'll be able to do a Show HN later this month. It's the best $1500 I've ever spent on anything related to software development or computers. You're not addressing the idea in the post, but the person. What the author achieved personally is irrelevant. What's important is that a very large and important industry is being completely transformed. Add: if I did it without LLMs it would have taken me a year, and would have been less complete.
- acedTrex 9mo ago
  "building a compiler that translates typescript to native binaries" Is a really weird way to say that you built a native compiler for typescript.
  - jeswin 9mo ago
    > Is a really weird way to say that you built a native compiler for typescript. Well, the Typescript team is rewriting their compiler (landing in v7) in go, and some ppl call it the native compiler. I think my statement is clearer? But then English is not my first language, so there's that.
- YouAreWRONGtoo 9mo ago
  [dead]

by ossa-ma 9mo ago

wg0 9mo ago

Well. TBH for some maybe this is the wrong question to ask but I have been thinking where did those 250 billion tokens go? What tools/products/services came out of that? EDIT: Typos

rocmcd 9mo ago
This has been my biggest question through this whole AI craze. If AI is making everyone X% more productive, where is the proof? Shouldn't we expect to see new startups now able to compete with large enterprises? Shouldn't we be seeing new amazing apps? New features being added? Shouldn't bugs be a thing of the past? Shouldn't uptime be a solved problem by now? I look around and everything seems to be... the same? Apart from the availability of these AI tools, what has meaningfully changed since 2020?
- Sevii 9mo ago
  AI coding improved a lot over 2025. In early 2025 LLMs still struggled with counting. Now they are capable of tool calling so they can just use a calculator. Frankly, I'd say AI coding may as well have not existed before mid-2025. The output wasn't really that good. Sure you could generate code but couldn't rely on a coding agent to make 2 line edits to a 1000 line file.
  - rocmcd 9mo ago
    I don't doubt that they have improved a lot this year, but the same claims were being made last year as well. And the year before that. I still haven't seen anything that proves to me that people are truly that much more productive. They certainly _feel_ more productive, though. Hell, the GP spent more than $50,000 this year on API calls alone and the results are... what again? Where is the innovation? Where are the tools that wouldn't have been possible to build pre-ChatGPT? I'm constantly reminded of the Feynman quote: "The first principle is that you must not fool yourself, and you are the easiest person to fool."

croemer 9mo ago

What's the source for the energy per token? I guess this? https://www.theguardian.com/technology/2025/aug/09/open-ai-chat-gpt5-energy-use?utm_source=chatgpt.com https://www.theguardian.com/technology/2025/aug/09/open-ai-c... 18Wh/1000t is on the high end of estimates. But even if it's 10x less I agree this is pretty crazy usage.

ossa-ma 9mo ago
"The University of Rhode Island based its report on its estimates that producing a medium-length, 1,000-token GPT-5 response can consume up to 40 watt-hours (Wh) of electricity, with an average just over 18.35 Wh, up from 2.12 Wh for GPT-4. This was higher than all other tested models, except for OpenAI's o3 (25.35 Wh) and Deepseek's R1 (20.90 Wh)." https://www.tomshardware.com/tech-industry/artificial-intelligence/chatgpt-5-power-consumption-could-be-as-much-as-eight-times-higher-than-gpt-4-research-institute-estimates-medium-sized-gpt-5-response-can-consume-up-to-40-watt-hours-of-electricity https://www.tomshardware.com/tech-industry/artificial-intell... https://app.powerbi.com/view?r=eyJrIjoiZjVmOTI0MmMtY2U2Mi00ZTE2LTk2MGYtY2ZjNDMzODZkMjlmIiwidCI6IjQyNmQyYThkLTljY2QtNDI1NS04OTNkLTA2ODZhMzJjMTY4ZCIsImMiOjF9 https://app.powerbi.com/view?r=eyJrIjoiZjVmOTI0MmMtY2U2Mi00Z... https://blog.samaltman.com/the-gentle-singularity https://blog.samaltman.com/the-gentle-singularity
- causal 9mo ago
  These numbers don't pass sanity check for me. With 4x300W cards you can get a 1K token DeepSeek R1 output in about 10 seconds. That's just 3.3Wh right? And that's before you even consider batching.

azan_ 9mo ago

That's very dishonest. Daily water intake is a fraction of how people use water! Producing 1 kg of beef requires 15 000 litres and if you put it that way (which is much more honest) it's not that bad. If you'd also take into account other ways people use water then it'd be even less shocking.

ossa-ma 9mo ago
And the other stats you ignored? As well as the main point I was making?
- azan_ 9mo ago
  I don't know about other stats, that's why I won't comment on them. But it doesn't matter - your water use stats are still manipulative and make your point much weaker.
  - ossa-ma 9mo ago
    I've adjusted the statistic for you: Adjusted beef consumption: 4.5 million litres of water can be used to produce 300kg of beef -> US (highest beef consumer/capita) consumes 23.3kg of beef , enough to feed ~13 Americans (30 Brits, ~43 Japanese) yummy delicious grass-fed beef yearly!
    - elevation 9mo ago
      These number seem off. A single head of cattle may contain 300kg of edible beef with a hanging weight of roughly twice that. In what world does raising a single bovine consume of 4.5 million liters of water? Neither the cow nor the cow's food retains much water; the water is merely delayed a little in its journey to the local watershed, and in vast parts of the US, local rainfall is adequate for this purpose (power irrigation isn't required for the crops, and cattle may drink from a pond.) Even if a cow drinks pumped well water, the majority of its nourishment will have been itself sustained by local natural rainfall. A datacenter's use of water over any timescale can hardly be compared with a cow's.

gwern 9mo ago

> So he's spent $51k on tokens in 3 months to build what exactly? Tools to enable you to spend more money on tokens? Sounds like that's more than a junior or an intern that would have cost twice as much in fully loaded cost.

aprilthird2021 9mo ago
He also put his own time into it. It's also more than hiring someone overseas esp just for a few months. Honestly it's more than most interns are paid for 3 months outside FAANG (considering housing is paid for there etc)
- gwern 9mo ago
  1. You put a lot of time into an intern or a junior too. 2. I didn't say 'paid', I said, 'fully loaded [total] cost'. The total cost of them goes far beyond their mere salary - the search process like all of the interviews for all candidates, onboarding, HR, taxes etc.
  - aprilthird2021 9mo ago
    1. Idk, I didn't have to. I managed an intern a few months back and he just did everything we had planned out and written down then started making his own additions on top. 2. Yeah I mentioned that also. 3. It's still more expensive than hiring a contractor esp abroad, even all in.

jeswin 9mo ago

I spent ~$1500 (and a bit more than two months) building a compiler that translates typescript to native binaries. I think I'll be able to do a Show HN later this month. It's the best $1500 I've ever spent on anything related to software development or computers. You're not addressing the idea in the post, but the person. What the author achieved personally is irrelevant. What's important is that a very large and important industry is being completely transformed. Add: if I did it without LLMs it would have taken me a year, and would have been less complete.

acedTrex 9mo ago
"building a compiler that translates typescript to native binaries" Is a really weird way to say that you built a native compiler for typescript.
- jeswin 9mo ago
  > Is a really weird way to say that you built a native compiler for typescript. Well, the Typescript team is rewriting their compiler (landing in v7) in go, and some ppl call it the native compiler. I think my statement is clearer? But then English is not my first language, so there's that.
YouAreWRONGtoo 9mo ago
[dead]