Chase AI
Region: US
Tuesday 11 November 2025 19:43:56 GMT
49126
1752
69
232
Music
Download
Comments
dosjshzhhff :
There is not a single AI product right now that isnt overhyped
2025-11-12 01:15:35
59
AlamIndah :
that’s rest api but with extra steps and cost
2025-11-12 05:39:50
59
Zain | Human Team :
This is helpful, in general I think AI has a last mile problem, easy to get started, hard to get right in the context of real use cases
2025-11-11 20:04:16
31
Dave C :
Yeah, I think that's part of why VS code copilot started using virtual tools that group together and consolidate tools that are attached to the agent to help reduce some of the token bloat. I've done some things similar with Langchain / Langgraph to only load in tools when they are actually needed and it helped a lot.
2025-12-29 05:51:53
1
seancon2000 :
Are we coming back
2025-12-28 18:06:51
0
I2C_Jason :
I’m telling you, we peaked with Pydantic, JSON, and GPT4o API calls sometime around late 2024.
2025-11-12 02:26:20
3
vanpancake77 :
If you thought they were more than an api, then yes you overdid it.
2025-11-12 03:40:36
8
Brian M :
So you're saying I need an MCP Server for my MCP Servers 😁
2025-11-12 04:13:42
1
plxnet :
maybe limit the tools it has access to begin with
2025-11-13 01:14:41
4
DMC :
LLM stuff overhyped??? No way??? I’m shocked!!
2025-11-12 12:41:22
3
Jon Capriola :
MCP isn’t secure at the protocol level I can spoof any agent in less than five lines of code.
2025-11-12 13:48:59
3
A Wild Pilot :
2/3?! That context window is basically useless
2025-11-11 21:14:13
4
Kaxf :
it sounds like there isn’t an alternative. regarding ur using tokens is it just how can u use the least amount with the correct context to execute the task?
2025-11-12 02:02:44
0
Bill :
AI is only a force multiplier at this point and cannot be relied upon in any way for mission critical business applications
2025-11-12 03:05:53
1
Nyan Helsing :
have the mcp server dynamically load endpoints based on context
2025-11-11 22:42:38
2
Nick :
Try adding HubSpot it will use up all tokens before it does anything
2025-11-11 22:33:29
1
Drain :
Anthropic: yo, we just solved AI’s contextual intelligence problem. Also Anthropic: yo, so uh, oops my bad
2025-11-12 09:59:08
1
wchar_t :
incremental discovery. it's fundamental
2026-01-04 10:00:32
0
Nidolab :
cloudflare is calling it "code mode"
2025-11-14 13:09:25
0
Vitja :
I have 12 Mcps using 800T - I have own System implemented working perfekctly
2025-12-04 12:41:42
0
888yea :
Can you give a link to this?
2026-01-12 17:23:40
0
Elusive Coffee :
MCP server for MCP servers 🙈😂
2025-12-22 04:29:29
0
Deadend Deafchild :
great video
2026-01-01 22:37:26
0
digitaltwin3 :
yes
2025-12-21 17:46:57
0
Duncan McKinnon :
This makes sense, and it seems like the complexity would be worth it because the agent will still use fewer tokens finding and using one or two tools than loading 100 fully formed.
2025-11-12 04:54:36
0
To see more videos from user @chase_ai_, please go to the Tikwm
homepage.
