ally krinsky | strategy & ai ✨
Region: US
Friday 15 May 2026 23:55:15 GMT
72267
3364
142
73
Music
Download
Comments
big_data_energy :
Craziest part… basic SQL is super easy to learn. SELECT columns FROM table WHERE logic ORDER by column LIMIT amount of rows… we’re reinventing SQL from first principles…
2026-06-24 02:00:18
76
homelander3310 :
plot twist users dont use it anyway
2026-05-16 14:29:55
318
Ash @ StockPixie :
Just get Metabase and sit it on top of/connect to your db. This is built in.
2026-07-01 18:55:45
0
wrsmith21 :
You still have the issue of the LLM generating the right query. Must have been one table and very simple questions. The more tables and the more columns means lower accuracy. Nobody has solved this problem at scale with consistent accuracy
2026-05-17 13:03:42
48
Nate :
You’re close, but you’re still not all the way there. You need a semantic layer tool like Strategy Mosaic connected with MCP.
2026-05-19 00:59:44
18
Anchal | Tech & Lifestyle :
Yes ai is a tool but it’s not the only tool or always the right tool
2026-05-16 19:29:42
81
elementone :
IMO Shouldn’t be hitting the db directly, should expose some sort of controllable interface / api in between along with defining skills to steer the ai to use your interface correctly. Allows you to do more role based gating, preventing destructive actions against the db, makes it easier to migrate to a different db/data store down the line
2026-05-17 15:28:28
12
Alyssa with AI | CPA :
We don’t need to build just to build. Over-engineering is one of the biggest risks (and time wasters)
2026-07-01 14:06:17
1
Uhhh :
Semantic models?
2026-06-18 01:22:27
3
socratestheturtle :
databricks genie spaces, next question
2026-05-16 19:24:45
8
nimra20143 :
What's the difference between in retrieving from the database than from the queries ? Why was the llm giving wrong numbers ?
2026-05-16 03:09:55
7
Joe :
Exactly. This is how Jira handles natural language queries. Just makes JQL
2026-05-23 04:19:52
3
birdly_68 ❌ :
Elon Musk has been on record as saying, engineering has tendency to over tune things that didn’t need to exist
2026-05-17 15:55:27
2
ajrodknee :
What does embedding the data do
2026-05-21 01:05:56
2
Rec0n :
A good tool call step up was all that’s needed
2026-05-18 02:08:44
0
Kaxf :
lol you had to give the the “ai” solution cause they asked for it but that wasn’t the best solution 😭
2026-05-16 01:24:47
18
cormorant :
lmao they were doing whattt
2026-05-28 05:40:19
2
hannah 🤸🏼♀️ :
stg 90% of use cases are teams wanting to ask questions over their files - sometimes i simply introduce them to a good ol ctrl F
2026-05-27 20:09:23
3
Dean Chianti :
Too much influence from IT
2026-05-17 13:23:45
2
John Brzozowski :
What’s a vector database
2026-05-17 12:28:12
3
Randall Smith :
God forbid anyone mention what the S in SQL stands for
2026-05-19 19:12:16
3
KDoubleArina :
Elasticsearch does this
2026-05-20 00:41:23
4
Chinchilla :
log the queries!!!
2026-05-17 18:40:43
2
S5505.0 :
Old video?
2026-05-18 02:33:09
2
heatculture :
Could your client not use snowflake?
2026-05-17 15:47:20
2
To see more videos from user @itsallykrinsky, please go to the Tikwm
homepage.
