Sammie

marketer. been in content, brand & performance. now mostly watching AI eat my toolkit one tool at a time.

[Sammie @nexgencoded]

i work in marketing. content, brand, performance - the whole mix.

in the last two years basically every tool i use has become 'AI-powered'

some of it is genuinely useful. a lot of it is just regular software with a chatbot stapled on and a price increase

this account is me figuring out which is which in real time. i'll tell you when something actually replaced a task i used to do.

i'll also tell you when a $99/month tool produced something i could've written in 10 minutes

honest takes. occasional rants. no affiliate links 😭

[Sammie @nexgencoded]

Webflow just launched Source.

and this is a much more interesting AI pitch than

we added a chatbot ✨

the idea is basically giving marketers AI coding-agent powers without first teaching them

git
branches
pull requests
why the build is suddenly broken😭

this is the kind of AI feature I actually want to test

[Sammie @nexgencoded]

@Gemini_Notebook flexible limits + 5 hour resets is such a better setup than come back tomorrow 😭 usage caps are way less annoying when they don’t kill the whole workday

[Sammie @nexgencoded]

one almond beating 38,000 ChatGPT queries in water use is the kind of stat that sounds fake until you remember how weird resource comparisons get

[Polymarket @Polymarket]

JUST IN: Sam Altman reveals a single California almond can require more water than 38,000 ChatGPT queries.

[Sammie @nexgencoded]

miley dropping the last name and entering the one-name club before a new album is such a power move

Cher
Madonna
Beyoncé
Miley

unfortunately my rebrand from person with deadlines to
Sammie has had less impact

https://t.co/6uvIkAu6b0

[Sammie @nexgencoded]

“Bass Persuades” already sounds like the kind of song that’s going to be impossible to avoid for a week straight lol

[Pop Base @PopBase]

MILEY teases her upcoming single ‘Bass Persuades.’

Out tomorrow.

[Sammie @nexgencoded]

@Dexerto GTA 6 literally saying “you’re beyond redemption” and turning off the morality system is insane 😭 rockstar really built a point of no return for being too evil

[Sammie @nexgencoded]

@ChrisGPT this case is going to matter way beyond OpenAI. if training on copyrighted text gets treated as broadly transformative, that changes the legal foundation for basically the whole model industry

[Sammie @nexgencoded]

AI spent the last few years learning to make pictures

now we’re giving it a camera and letting it move around inside them

totally normal progression. nothing terrifying about that sentence whatsoever

[Sammie @nexgencoded]

AI tool founders need to stop naming features like we're entering battle

Agent Mode

Deep Research

Operator

Mission Control

brother I asked it to summarize a PDF 😭

what's the most unnecessarily dramatic AI feature name you've seen??

[Sammie @nexgencoded]

I've tested enough AI tools to realize there are basically 3 categories

1. genuinely changes how you work

2. useful, but wildly overpriced

3. regular software with a chatbot stapled on

I'm trying to figure out which is which so you don't have to

one teardown every week on Saturday

[Sammie @nexgencoded]

one question I wish every AI tool had to answer

what can you do that ChatGPT + 10 minutes can't??

if the answer is

we have a nicer dashboard....

I'm keeping my $29/month 😭

genuinely becoming my first test before adding anything to the stack

[Sammie @nexgencoded]

every AI tool homepage

save 10+ hours every week

me after trying it

45 mins learning the tool

20 mins fixing the output

15 mins wondering if I could've just done it myself

time saved?? spiritually, maybe....

[Sammie @nexgencoded]

@PopBase @miley different religion → let’s get married as tracks 2 and 3 is already telling me this album is going to be emotionally irresponsible 😭

miley please i have work this month

[Sammie @nexgencoded]

@claudeai the model release cadence is getting ridiculous 😭

but best for coding + knowledge work is exactly where the real test gets interesting

benchmarks are nice

give it a messy 2-hour task and see if it still knows what you asked at minute 47

[Sammie @nexgencoded]

Anthropic just dropped Claude Fable 5.1.

apparently we're already doing .1 updates to frontier models now 😭

Anthropic is positioning it around coding, knowledge work + research

the interesting part isn't another benchmark bump though

it's whether 5.1 is actually noticeably better on the annoying long tasks where Claude starts strong and then slowly forgets what planet it's on

gonna be watching real workflows way more than leaderboard screenshots

[Sammie @nexgencoded]

Salesforce + Claude is getting more interesting

not because companies desperately needed another chatbot

but because Claude can increasingly sit on top of the stuff Salesforce already knows-

customers

pipeline

support

workflows

that's the AI product pattern I'm watching now

the model isn't the moat

the useful context underneath it might be

[Sammie @nexgencoded]

@Da7_Tech google’s biggest problem might just be perception lag. once people decide a model is “mid,” it takes way longer than one good release to change that

[Sammie @nexgencoded]

me- I should probably cancel some AI subscriptions

also me 6 minutes later-

wait this one turns PDFs into podcasts AND has a chrome extension??

$19/month

added to stack...🫣

financial discipline resumes Monday

[Sammie @nexgencoded]

this is the real benchmark most users care about: does it sound normal, get to the point, and cost less to run

[Yuchen Jin @Yuchenj_UW]

I barely use Claude Opus 5 these days.

It doesn’t talk like a normal person (especially in Chinese). Way too verbose and rarely to the point. GPT-5.6 Sol is better.

Meanwhile, OSS models like GLM-5.3, GLM-5.3-Flash, and Kimi K3 are pretty good and cost-efficient.

There’s no default winner anymore.

[Sammie @nexgencoded]

@abacaj this is why model loyalty makes less sense every month. if the output holds and the bill drops, switching should be boring