zainhoda

6 ms·

1.
by zainhoda 7mo ago
Working on a web framework that provides some guardrails around what a coding agent can and can’t touch without human approval. Makes it easier to have confidence in 5000 line code changes without having to comb through the code. https:&#x
2.
Is your SaaS going to survive?(zainhoda.substack.com)
4 points by zainhoda 8mo ago|0 comments
3.
by zainhoda 1y ago
It would be amazing if the showcase was itself vibe coded. If so, you should showcase it!
4.
by zainhoda 1y ago
Mobile app that lets you continue coding while you’re away from your computer. The goal is to be a full mobile IDE that lets you use Claude Code, Gemini CLI, and other agentic code editors. Has mobile-native file browsing and git integratio
5.
Vibe Coding CLI(github.com)
2 points by zainhoda 1y ago|0 comments
6.
by zainhoda 1y ago
Oof… coincidentally my task for today is to implement Sign in with Apple. If you were starting over what would you do differently?
7.
by zainhoda 2y ago
What's The State Of Elm? I love Elm but I think it's pretty clear that Evan is effectively declaring it as abandonware because he couldn't figure out a business model to sustain him. What’s Evan Working On? Sounds like he&#x2
8.
by zainhoda 2y ago
I’m waiting for the same signal. There are essentially 2 vastly different states of the world depending on whether GPT-5 is an incremental change vs a step change compared to GPT-4.
9.
by zainhoda 2y ago
Would you be interested in merging with Vanna in some way? You’re ahead of us in terms of interface but we’re ahead of you in terms of adoption (because of specific choices we’ve made and partnerships we’ve done).
10.
by zainhoda 2y ago
Nice job getting something released! How does this compare to the other similar open source solutions like Vanna AI and DataHerald?
11.
Build a chatbot for your Azure SQL database(devblogs.microsoft.com)
1 points by zainhoda 2y ago|0 comments
12.
by zainhoda 2y ago
Looks really interesting! I saw something in the code about streaming. Could you explain that a bit more?
13.
by zainhoda 2y ago
I've been looking for something like this! Does it optimize the prompt template for LangChain only or is there a way I can get it to generate a raw system prompt that I can pass to the OpenAI API directly?
14.
by zainhoda 2y ago
I think of all the options for “carrots” and “sticks” that companies are offering for RTO, I like this one the best.
15.
by zainhoda 3y ago
Very cool. Would the license allow for use with Vanna? https://github.com/vanna-ai/vanna
16.
by zainhoda 3y ago
Founder of Vanna AI here -- appreciate the link and I agree, we're solving slightly different problems.
17.
by zainhoda 3y ago
I’m trying to solve for this with my project using RAG and (at least based on what people say in Discord), it’s working really well for them: https://github.com/vanna-ai/vanna
18.
by zainhoda 3y ago
That's a fair concern. In actual usage, however, the vast majority of hallucination (>90%) tends to be: - phantom tables and columns, in which case the query will fail - incorrect syntax for date functions (i.e. the wrong flavor of
19.
by zainhoda 3y ago
Author of the package here. Apologies if this wasn't clear in the documentation, but what you're talking about is absolutely possible. Feel free to join our Discord -- we have other users who have this multi-tenant setup.
20.
by zainhoda 3y ago
Yes, 100%! Can you turn this comment into a blog post so that I can send it to people who make this claim?
21.
by zainhoda 3y ago
One added benefit of RAG is that it's more "pluggable." It's a lot easier to plug into newer LLMs that come out. If and when GPT-5 comes out, it'll be a one character change in your code to start using it and still
22.
by zainhoda 3y ago
I've noticed this as well. Do you have any theories as to why that is?
23.
by zainhoda 3y ago
What would be missing from that is the ability to look up what tables are in the database, how they join together, what terminology the business uses, preferences around whether to exclude nulls from certain columns, etc. This allows you to
24.
by zainhoda 3y ago
That's basically what happens but the power is that in Python, you can do this: sql = vn.generate_sql(question=...) Which means that now the SQL can be executed and you can get the table, chart, etc in any interface. Flask: https:&#x2
25.
by zainhoda 3y ago
Super interesting! Why did you decide to shut down?
26.
by zainhoda 3y ago
Author of the package here. Thank you! That's exactly what we've seen. Businesses spent millions of dollars getting all their structured data into a data warehouse and then it just sits there because the people who need the inform
27.
by zainhoda 3y ago
100% -- in fact originally the package used to parse out SELECT statements and only execute SELECT statements. After some feedback, we decided that the permissions on the user should handle that level of detail.
28.
by zainhoda 3y ago
Author of the package here. I think that it probably could handle that but may need more example SQL queries.
29.
by zainhoda 3y ago
Author of the package here. Joining 5 tables is not a problem. The training does not necessarily require you to write the queries by hand. A trick that we've seen people do is to just train with DDL statements and then ask "leadin
30.
by zainhoda 3y ago
Author of the package here. Yes, this is a very common use case.

More ›