J.O.
@radio46.net
If you are on VPN, you usually don’t select Nigeria or Bangladesh as exit nodes.
November 23, 2025 at 9:53 PM
If you are on VPN, you usually don’t select Nigeria or Bangladesh as exit nodes.
Levels of reasoning impact this big time, too. Asking “How are you?”: Low reasoning: 50 tokens. Medium reasoning: 81 tokens. High reasoning: 119 tokens. Model with no reasoning: 33 tokens.
October 15, 2025 at 2:59 AM
Levels of reasoning impact this big time, too. Asking “How are you?”: Low reasoning: 50 tokens. Medium reasoning: 81 tokens. High reasoning: 119 tokens. Model with no reasoning: 33 tokens.
Some question, not same output. But that’s part of the problem, as models “reason” more and provide supposedly more nuanced answers to the same quesions, they burn more tokens.
October 15, 2025 at 1:13 AM
Some question, not same output. But that’s part of the problem, as models “reason” more and provide supposedly more nuanced answers to the same quesions, they burn more tokens.
Playing with local LLMs provides an interesting anecdotal insight into this. Asking a reasoning model "How are you?" burns 40-50 tokens. Non-reasoning, older model burns ~20. Asking to explain Pythagoras theorem burns 1000 tokens with a newer model vs 250 tokens with an older model.
October 15, 2025 at 12:01 AM
Playing with local LLMs provides an interesting anecdotal insight into this. Asking a reasoning model "How are you?" burns 40-50 tokens. Non-reasoning, older model burns ~20. Asking to explain Pythagoras theorem burns 1000 tokens with a newer model vs 250 tokens with an older model.
This album unlocked music for me. Thank you.
October 5, 2025 at 1:18 PM
This album unlocked music for me. Thank you.
What about companies like Adobe who have built own models as a complement to their core business, without over-leveraging on compute?
September 19, 2025 at 2:42 AM
What about companies like Adobe who have built own models as a complement to their core business, without over-leveraging on compute?
Genuine question — don’t you see some demand in design, graphics video space? Not as an creative tool, more to speed up and automate workflows (video editing, image manipulation). Marketing, advertising mostly.
September 19, 2025 at 1:46 AM
Genuine question — don’t you see some demand in design, graphics video space? Not as an creative tool, more to speed up and automate workflows (video editing, image manipulation). Marketing, advertising mostly.
Adobe's still making it very clear:
September 18, 2025 at 12:40 AM
Adobe's still making it very clear:
Yes, and as that article itself says, it’s been clarified. Adobe does not train AI on customer data - that’s actually one of the reasons why quality of Adobe’s image generation tools is lower as it’s trained on a much smaller dataset of mostly Adobe Stock images.
September 7, 2025 at 12:09 AM
Yes, and as that article itself says, it’s been clarified. Adobe does not train AI on customer data - that’s actually one of the reasons why quality of Adobe’s image generation tools is lower as it’s trained on a much smaller dataset of mostly Adobe Stock images.
“Adobe training AI on your media” is a conspiracy theory. It’s simply not true. Be better.
September 4, 2025 at 9:10 PM
“Adobe training AI on your media” is a conspiracy theory. It’s simply not true. Be better.
Good ad placement on Truth Social
September 2, 2025 at 3:57 AM
Good ad placement on Truth Social
You mention ChatGPT MAU in the article: I only have access to mobile app data, but the app MAU is approx 550M (up from ~200M a year ago), DAU ~250M, WAU ~500M. App store revenue ~$1.5bn in past year (growing). Usage churn is <10%, which is quite solid.
August 26, 2025 at 11:20 AM
You mention ChatGPT MAU in the article: I only have access to mobile app data, but the app MAU is approx 550M (up from ~200M a year ago), DAU ~250M, WAU ~500M. App store revenue ~$1.5bn in past year (growing). Usage churn is <10%, which is quite solid.
Is there a how to on setup and installation of Tools?
August 12, 2025 at 8:00 AM
Is there a how to on setup and installation of Tools?
I'm aware the model type is 'embedded' in the model itself and its training data. Just found it interesting that gpt-5-nano consistent calls out it's from the GPT-4 family, and I was wondering if there is some routing happening behind the scenes, or if the mini/nano models are simply rebadged GPT-4.
August 8, 2025 at 11:33 PM
I'm aware the model type is 'embedded' in the model itself and its training data. Just found it interesting that gpt-5-nano consistent calls out it's from the GPT-4 family, and I was wondering if there is some routing happening behind the scenes, or if the mini/nano models are simply rebadged GPT-4.
speaking of API, it looks like "gpt-5-nano" (called via API) self-identifies as GPT-4, unlike whatever model is available on chatgpt.com. Not sure how much I should read into this.
August 8, 2025 at 10:53 PM
speaking of API, it looks like "gpt-5-nano" (called via API) self-identifies as GPT-4, unlike whatever model is available on chatgpt.com. Not sure how much I should read into this.
The screenshot shows Photoshop Express, which is a different (older) app.
July 5, 2025 at 1:30 PM
The screenshot shows Photoshop Express, which is a different (older) app.
Create a second account for trading
June 30, 2025 at 10:18 PM
Create a second account for trading
I'm not seeing it - the link directs me to standard Inbox view.
June 26, 2025 at 8:21 AM
I'm not seeing it - the link directs me to standard Inbox view.
How is this relevant though? Otto Abetz died when Eric Abetz was 4 months old.
June 9, 2025 at 4:38 AM
How is this relevant though? Otto Abetz died when Eric Abetz was 4 months old.
“Following this and an early season run of six defeats and two draws in eight games, a banner declaring, "Three years of excuses and it's still crap ... ta-ra Fergie." was displayed at Old Trafford, and many journalists and supporters called for Ferguson to be sacked.”
Reminds me of this.
Reminds me of this.
June 7, 2025 at 12:00 AM
“Following this and an early season run of six defeats and two draws in eight games, a banner declaring, "Three years of excuses and it's still crap ... ta-ra Fergie." was displayed at Old Trafford, and many journalists and supporters called for Ferguson to be sacked.”
Reminds me of this.
Reminds me of this.
Same story - needs a beta firmware update which can be downloaded from the same links that are circulating online. Updated mine and it works now.
June 6, 2025 at 10:05 AM
Same story - needs a beta firmware update which can be downloaded from the same links that are circulating online. Updated mine and it works now.