Sherveen
@sherveen.bsky.social
The squeeze is worth it, even if there's no juice.
Founder @ Free Agency.
Build your AI fluency → http://aimuscle.com
For those who notice nuance → http://noticethenuance.com
YouTube (tech, AI, venture, society) → http://youtube.com/@sherveenshow
Founder @ Free Agency.
Build your AI fluency → http://aimuscle.com
For those who notice nuance → http://noticethenuance.com
YouTube (tech, AI, venture, society) → http://youtube.com/@sherveenshow
Not ready to call it on this one yet, but...
GPT-5 Pro might be better than GPT-5.1 Pro. They're quite different.
The 5.1 flavor is definitely more directed, more "narrow" on the prompt it was given. And one of 5 Pro's strengths is in how it 'explores' a problem space.
GPT-5 Pro might be better than GPT-5.1 Pro. They're quite different.
The 5.1 flavor is definitely more directed, more "narrow" on the prompt it was given. And one of 5 Pro's strengths is in how it 'explores' a problem space.
November 21, 2025 at 6:46 PM
Not ready to call it on this one yet, but...
GPT-5 Pro might be better than GPT-5.1 Pro. They're quite different.
The 5.1 flavor is definitely more directed, more "narrow" on the prompt it was given. And one of 5 Pro's strengths is in how it 'explores' a problem space.
GPT-5 Pro might be better than GPT-5.1 Pro. They're quite different.
The 5.1 flavor is definitely more directed, more "narrow" on the prompt it was given. And one of 5 Pro's strengths is in how it 'explores' a problem space.
OK, deep breath...
Grok 4.1 Thinking is a model worth its effort.
I don't know where to put it yet, but it's a solid #3 or #4, whereas Grok 4 (& Heavy) were a very distant 5 or 6, and were _rapidly_ degrading over time for some reason.
Grok 4.1 Thinking is a model worth its effort.
I don't know where to put it yet, but it's a solid #3 or #4, whereas Grok 4 (& Heavy) were a very distant 5 or 6, and were _rapidly_ degrading over time for some reason.
November 20, 2025 at 3:57 PM
OK, deep breath...
Grok 4.1 Thinking is a model worth its effort.
I don't know where to put it yet, but it's a solid #3 or #4, whereas Grok 4 (& Heavy) were a very distant 5 or 6, and were _rapidly_ degrading over time for some reason.
Grok 4.1 Thinking is a model worth its effort.
I don't know where to put it yet, but it's a solid #3 or #4, whereas Grok 4 (& Heavy) were a very distant 5 or 6, and were _rapidly_ degrading over time for some reason.
Nano Banana Pro is quite good.
Prompt: "screenshot of twitchtv, someone playing Star Wars: Battlefront 4, their face in the bottom right corner, twitch in dark mode, chat going crazy with emojis"
Comparisons to a few other image models, 👇.
Prompt: "screenshot of twitchtv, someone playing Star Wars: Battlefront 4, their face in the bottom right corner, twitch in dark mode, chat going crazy with emojis"
Comparisons to a few other image models, 👇.
November 20, 2025 at 3:47 PM
Nano Banana Pro is quite good.
Prompt: "screenshot of twitchtv, someone playing Star Wars: Battlefront 4, their face in the bottom right corner, twitch in dark mode, chat going crazy with emojis"
Comparisons to a few other image models, 👇.
Prompt: "screenshot of twitchtv, someone playing Star Wars: Battlefront 4, their face in the bottom right corner, twitch in dark mode, chat going crazy with emojis"
Comparisons to a few other image models, 👇.
Gemini 3 Pro has some weird quirks with tool calling, even in Google's own environments.
Ex. it'll often type out code changes to me in chat in AI Studio and think that it updated the file. I then have to nudge it to make the actual change.
Ex. it'll often type out code changes to me in chat in AI Studio and think that it updated the file. I then have to nudge it to make the actual change.
November 20, 2025 at 3:22 PM
Gemini 3 Pro has some weird quirks with tool calling, even in Google's own environments.
Ex. it'll often type out code changes to me in chat in AI Studio and think that it updated the file. I then have to nudge it to make the actual change.
Ex. it'll often type out code changes to me in chat in AI Studio and think that it updated the file. I then have to nudge it to make the actual change.
I'm so confused. Why is he torching his reputation with the worst timing possible?
We need VCs to fund an infinity amount of Salesforce replacements until we can figure out what he hell is going on.
We need VCs to fund an infinity amount of Salesforce replacements until we can figure out what he hell is going on.
November 20, 2025 at 2:56 PM
I'm so confused. Why is he torching his reputation with the worst timing possible?
We need VCs to fund an infinity amount of Salesforce replacements until we can figure out what he hell is going on.
We need VCs to fund an infinity amount of Salesforce replacements until we can figure out what he hell is going on.
The _style_ of this robot (Memo) is underrated. Things like NEO and Figure look... _cool_, but...
Memo looks more like Matic (smart home vacuum) -- in a good way! It's like you took something I'd already buy for my home, and made it 100x better, rather than 'roboticizing' a human.
Memo looks more like Matic (smart home vacuum) -- in a good way! It's like you took something I'd already buy for my home, and made it 100x better, rather than 'roboticizing' a human.
November 20, 2025 at 12:32 AM
The _style_ of this robot (Memo) is underrated. Things like NEO and Figure look... _cool_, but...
Memo looks more like Matic (smart home vacuum) -- in a good way! It's like you took something I'd already buy for my home, and made it 100x better, rather than 'roboticizing' a human.
Memo looks more like Matic (smart home vacuum) -- in a good way! It's like you took something I'd already buy for my home, and made it 100x better, rather than 'roboticizing' a human.
Fascinating bit of pricing and monetization economics here from Travis.
Immediately made me want more of this sort of product and market theory content from founders.
Immediately made me want more of this sort of product and market theory content from founders.
November 19, 2025 at 11:55 PM
Fascinating bit of pricing and monetization economics here from Travis.
Immediately made me want more of this sort of product and market theory content from founders.
Immediately made me want more of this sort of product and market theory content from founders.
I genuinely want the next few to be...
gpt-5.1-codex-max-harder
gpt-5.1-codex-max-better
gpt-5.1-codex-max-harder-better
then
gpt-5.1-codex-max-faster
gpt-5.1-codex-max-stronger
gpt-5.1-codex-max-harder-better-faster
then
gpt-5.1-codex-max-harder-better-faster-stronger
gpt-5.1-codex-max-harder
gpt-5.1-codex-max-better
gpt-5.1-codex-max-harder-better
then
gpt-5.1-codex-max-faster
gpt-5.1-codex-max-stronger
gpt-5.1-codex-max-harder-better-faster
then
gpt-5.1-codex-max-harder-better-faster-stronger
November 19, 2025 at 8:55 PM
I genuinely want the next few to be...
gpt-5.1-codex-max-harder
gpt-5.1-codex-max-better
gpt-5.1-codex-max-harder-better
then
gpt-5.1-codex-max-faster
gpt-5.1-codex-max-stronger
gpt-5.1-codex-max-harder-better-faster
then
gpt-5.1-codex-max-harder-better-faster-stronger
gpt-5.1-codex-max-harder
gpt-5.1-codex-max-better
gpt-5.1-codex-max-harder-better
then
gpt-5.1-codex-max-faster
gpt-5.1-codex-max-stronger
gpt-5.1-codex-max-harder-better-faster
then
gpt-5.1-codex-max-harder-better-faster-stronger
Gemini's new Agent mode has a neat canvas-style UI for acting inside your Gmail inbox.
Still doesn't seem to successfully adopt user writing styles in drafting the emails themselves, but neat nonetheless.
Still doesn't seem to successfully adopt user writing styles in drafting the emails themselves, but neat nonetheless.
November 19, 2025 at 12:27 AM
Gemini's new Agent mode has a neat canvas-style UI for acting inside your Gmail inbox.
Still doesn't seem to successfully adopt user writing styles in drafting the emails themselves, but neat nonetheless.
Still doesn't seem to successfully adopt user writing styles in drafting the emails themselves, but neat nonetheless.
Google's AI Mode w/ Gemini 3 gets fast, easy generative UI as part of its elevated search experience.
I think this is really underrated.
I think this is really underrated.
November 18, 2025 at 7:29 PM
Google's AI Mode w/ Gemini 3 gets fast, easy generative UI as part of its elevated search experience.
I think this is really underrated.
I think this is really underrated.
This is unbelievably funny and real.
November 18, 2025 at 6:20 PM
This is unbelievably funny and real.
The era of generative UI begins.
"As an example of how we are building on top of Gemini 3, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, all generated completely on the fly based on your query."
"As an example of how we are building on top of Gemini 3, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, all generated completely on the fly based on your query."
November 18, 2025 at 5:10 PM
The era of generative UI begins.
"As an example of how we are building on top of Gemini 3, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, all generated completely on the fly based on your query."
"As an example of how we are building on top of Gemini 3, AI Mode in Search now uses Gemini 3 to enable new generative UI experiences, all generated completely on the fly based on your query."
Wow @ Gemini 3 Deep Think.
November 18, 2025 at 4:40 PM
Wow @ Gemini 3 Deep Think.
Gemini 3 Pro benchmarks are out:
November 18, 2025 at 11:17 AM
Gemini 3 Pro benchmarks are out:
I now spend the dark hours w/ AI building personal software that will compound my context engineering.
Have been doing so throughout 2025.
Ya cannot fathom.
Have been doing so throughout 2025.
Ya cannot fathom.
November 18, 2025 at 7:54 AM
I now spend the dark hours w/ AI building personal software that will compound my context engineering.
Have been doing so throughout 2025.
Ya cannot fathom.
Have been doing so throughout 2025.
Ya cannot fathom.
Say what you will about Olivia Nuzzi.
She romanticizes American politics more than the rest of us.
A true patriot.
She romanticizes American politics more than the rest of us.
A true patriot.
November 18, 2025 at 6:57 AM
Say what you will about Olivia Nuzzi.
She romanticizes American politics more than the rest of us.
A true patriot.
She romanticizes American politics more than the rest of us.
A true patriot.
Everyone wants to talk about Gemini 3.0, but I want to talk about @GoogleDeepMind's SIMA 2.
An embodied agent (with reasoning powered by Gemini) that's getting really good at playing video games and generalizing concepts across games.
And then you toss it into...
An embodied agent (with reasoning powered by Gemini) that's getting really good at playing video games and generalizing concepts across games.
And then you toss it into...
November 17, 2025 at 6:14 PM
Everyone wants to talk about Gemini 3.0, but I want to talk about @GoogleDeepMind's SIMA 2.
An embodied agent (with reasoning powered by Gemini) that's getting really good at playing video games and generalizing concepts across games.
And then you toss it into...
An embodied agent (with reasoning powered by Gemini) that's getting really good at playing video games and generalizing concepts across games.
And then you toss it into...
How do we solve the misalignment problems within humans?
November 17, 2025 at 1:52 PM
How do we solve the misalignment problems within humans?
Reid going on an absolute once-in-a-lifetime "fuck it, fuck you" tweet run. Love to see it.
I have no idea who is in the files, but the way Reid's former friends, colleagues, and partners on the tech right have treated him over the past several years has been grotesque.
I have no idea who is in the files, but the way Reid's former friends, colleagues, and partners on the tech right have treated him over the past several years has been grotesque.
November 15, 2025 at 4:13 AM
Reid going on an absolute once-in-a-lifetime "fuck it, fuck you" tweet run. Love to see it.
I have no idea who is in the files, but the way Reid's former friends, colleagues, and partners on the tech right have treated him over the past several years has been grotesque.
I have no idea who is in the files, but the way Reid's former friends, colleagues, and partners on the tech right have treated him over the past several years has been grotesque.
Tech Twitter is kind of turning on Marc Andreessen tonight.
There may be hope yet, everyone.
There may be hope yet, everyone.
November 9, 2025 at 4:29 AM
Tech Twitter is kind of turning on Marc Andreessen tonight.
There may be hope yet, everyone.
There may be hope yet, everyone.
The dude who receives personalized healthcare from the manosphere oligarchs has something to say.
Adjacency is complacency. If you support this admin, you are an enabler.
Adjacency is complacency. If you support this admin, you are an enabler.
November 5, 2025 at 3:27 PM
The dude who receives personalized healthcare from the manosphere oligarchs has something to say.
Adjacency is complacency. If you support this admin, you are an enabler.
Adjacency is complacency. If you support this admin, you are an enabler.