Citi says customers granting third-party AI agents account access are responsible for their activity. I'd separate reading finances from moving money. Check permissions and your bank's terms before handing an assistant the keys. A useful tool doesn't need unlimited authority.
2
14
The Freez Memo is where I write about AI decisions owners actually have to make: costs, permissions and what to keep under human control. Subscribe: freez.digital/memo
5
I run Freez Digital and our own AI lab. Here I share what we're building, what breaks, and what business owners can actually use. Not every task needs AI. Not every model needs to run locally. The useful question: does it make the work better?
1
12
I go deeper in the Freez Memo: freez.digital/memo
6
I don't care how cheap an AI answer is if someone has to redo the work. When comparing models, count the bill, retries and human fixes. Same task. Same standard. Make the model earn the switch on your work, not the vendor's demo.
2
34
Alibaba's release describes its own benchmark, not a test of your workflow. Source: alibabacloud.com/blog/alibab…
58
A useful introduction starts with a problem, not a pitch. I'd like to meet owners of practices, firms or crews with one repetitive job they want an honest answer on. Ask them first. Who was the last person you met through an introduction that actually helped?
1
17
Chase Freezman retweeted
GPT has gotten so slow I am so grateful for the hardware I own and DeepSeek v4.1 Flash 1017
Having thousands of tokens per second is something I don’t think I can live without anymore
Louder for the people in the back!! Or the Agents in the back. Give your Agents as many tools as you can!
SHOCKING: You can use both local and cloud models!
Seems like people forgot this is actually possible. Claiming one has no advantage over the other misses the point.
They have different strengths for different use cases. If you're smart about it, you should use both!
2
102
Seems like GPT 5.6 Sol is flying through my usage again (and my timelines). Same workload as last week but seeing a significant decrease in usage. Still very efficient but this roller coaster is wild!
1
53
When the family goes to sleep tonight. I’m loading Deep Seek on DGX spark to finally put it to work! Excited to test this out. 😃 🤖
3
93
Im seeing double! Can’t wait to start testing these after client calls this afternoon! #DGXSpark #Nvidia #FreezDigital
2
61
I just got 2 DGX Sparks from Nvidia at Microcenter and when searching on X I see a ton of posts from amazing contributors that are all running pretty significant local tests which is pretty remarkable but I use mine to power and run my entire marketing agency…
1
2
46






