Everyone is a programmer

Everyone is a programmer

This statement was made by the CEO of Nvdia back in May of 2023. I dismissed that statement as hyperbole. Back then ChatGPT was still new and I had played around with it a little but apart from generating haiku's and asking it to explain relativity to me like a 5 year old, I hadn't done much.

Fast forward 18 months and this statement seems to be more real to me. To me being a programmer has always meant writing code and despite working in Tech for 20+ years, I have mostly stayed on the GTM side and away from the programming side. So this statement was just a statement. Why would I need to program.

Then this weekend I signed up for ChatGPT's $200 per month offering as I had gnarly problem I was trying to solve. I started my own LLC back in Feb and I know that there are decisions I make that have tax implications that can positively or negatively affect me (in the thousands). The obvious answer is to get an accountant and then I realized that accountants will do your taxes, maybe give you some advice but they are often reacting to information you give them. But if you don't know what you don't know and you don't give them information, they may or may not ask the questions that can get the information from you.

Here are a series of such questions -

  1. LLC vs S Corp - Which saves me more in taxes? What if I got a severance and I have paid most of my payroll taxes, and that is not a big factor?
  2. Am I eligible for R&D tax credit - Most people think that because I am in consulting I am not eligible and so don't bother to ask, but it also applies to people that drive productivity.
  3. Solo 401K or a SEP IRA - Also, should I max out my 401K? Which leads to a lower tax outlay? Which leads to more cash flow? What is the right balance?
  4. As an S Corp how much should I take in salary? If my salary is eligible for R&D tax credit, should I take a higher salary? But then how does this affect QBI?
  5. Can I rent my house to myself for quarterly board meetings?

I needed to run a bunch of scenarios and come up with what the optimal salary to pay myself, how much to put in my 401K and how much to have my S Corp contribute to my 401K. I needed to optimize this so I could maximize my take home pay + investments while minimizing taxes.

No accountant was going to help me do this, or if they did they would charge a lot of money for this. I had someone who wanted $10K to be this advisor.

So I spent this weekend working with GPT o1 pro and had it build me an app that would take in my Gross Business Revenue and my business expenses (not including Employer payroll taxes, not including Employer 401K contributions, not including salary) and return an optimal salary for me. I gave it constraints - salary had to be at least X, business could not show a loss, I could not leave any unused R&D tax credit.

I did this all by myself in a few hours. I am not a programmer, and I don't know how to write or deploy programs. I used to be one 20 years ago but that has long been forgotten. GPT o1 Pro wrote the code, told me how to deploy it on Netlify and then the biggest eye opener was the debugging and iterating. Every time I found something off in the output while testing, I would just tell it to make changes in English and it would regenerate the code.

BTW.. I gave the same instructions to Claude, Gemini and GPT 4o and they all struggled generating something that worked. So in that sense the $200 is well worth the investment if you have such needs.

This is a great example of the kind of programming that I would never do and I would never pay someone to do (at least not a programmer who would charge tens of thousands).

These are the kinds of prompts I was giving it -

Remember that the intent of this application is to maximizing the take home money. The inputs are Gross Revenue and Expenses only. Using this it should run scenarios on salary, employee 401K contribution, employer 401K contribution. It should use R&D tax credit as a constant at 14% of Salary. Employer 401k contribution cannot be greater than 25% of salary. Employee 401K contribution cannot be greater than 30,500. Assume QBI is 20% of business profit. Use hypothetical standard deduction as 15000 and hypothetical tax brackets as.- 
 10%: $0 - $11,500, 12%: $11,501 - $45,000,  22%: $45,001 - $97,000, 24%: $97,001 - $185,000. The output should be the take home pay, the salary, the employer 401K contribution, the employee 401K contribution, payroll taxes, federal income tax. There should also be detailed calculations explaining everything        

This is how I was providing change requests

Remember that payroll taxes only apply for salary upto 160000. Also, one constraint is that the business cannot make a loss.        
One clairification on this statement - Another constraint is to make sure that there is no unused R&D tax credit. Either the R&D tax credit is used to completely offset the federal taxes or R&D tax credit is less than total federal taxes owed. It should actually be that the R&D tax credit is used to completely offset the federal taxes + employer portion of payroll taxes or R&D tax credit is less than federal taxes + employer portion of payroll taxes        

Everytime it generated updated code that I could then save and test. Here is the app if anyone wants to play with it - https://tiny-jelly-9bc86e.netlify.app/


As I was thinking about this I realized that I have been a programmer for the last 6 months. All the projects that I have done for my clients around GenAI have required me to build agents that do some work and provide some output. What I realized is that the job of a programmer is not to write programs, it is to generate software that solves a business need. This is what an agent is and I have been using low code / no code tools like Anyquest.ai to do this. I've built agents that research executives, companies, read through case studies and pull out relevant snippets etc.

Most recently, I have been playing around with Agent.ai and built a few different agents like -

  • Executive DISC profile - It takes an executives name and generates their DISC profile and gives you tips on how to communicate with them.
  • Fit to purpose image generator - It takes specific inputs to generate the image you want.
  • Influencer social media summarizer - It takes the name of an influencer and summarizes their Tweets, LinkedIn posts, YouTube videos and latest news. In case you don't have time to keep up with them real time, this helps you get a monthly digest.

Without knowing it I have become a programmer!!

Jeff Thomson

Owner, Thomson Engineering, Inc.

1 个月

Coming to a PC near you.

Jeff Huckaby

CEO and Co-Founder | Passionate about helping people have better analytics outcomes using consulting, talent acquisition, and analytics solutions as a service.

2 个月

The world is changing fast.

回复
Mike Walsh

Struck by lightning in a cemetery when I was 14. Founding Partner at Structure Capital - seed investor in Salesforce, Uber, Sonder, Carta, Peerspace, Forge, HiFI Labs, OhmConnect, Zack and Lucas

3 个月

I would love to see what you're building.

This post is an excellent example of "services as software." You got accounting services delivered to you by software, which was developed for you by o1, yet another piece of software. It's software all the way down. Professional services is a 6.5-trillion dollar industry ... just saying.

Nice work! And an inspiring description of how to get things done.

要查看或添加评论,请登录

Vikram Ekambaram的更多文章

  • Vibe Prompting and Agent Debugging

    Vibe Prompting and Agent Debugging

    Vibe Prompting - A Two Part Series is sponsored by Agent.ai - Discover, connect with and hire AI agents to do useful…

    9 条评论
  • Agent Building and Vibe Prompting - #1

    Agent Building and Vibe Prompting - #1

    Vibe Prompting - A Two Part Series is sponsored by Agent.ai - Discover, connect with and hire AI agents to do useful…

    3 条评论
  • A person walks onto your website...

    A person walks onto your website...

    Mixture of Agents (MOAT) - A Two Part Series is sponsored by Agent.ai - Discover, connect with and hire AI agents to do…

    4 条评论
  • The Future is Multi-Agentic

    The Future is Multi-Agentic

    Mixture of Agents (MOAT) - A Two Part Series is sponsored by Agent.ai - Discover, connect with and hire AI agents to do…

    1 条评论
  • Anthropic Agentic Systems - #5. Evaluator-Optimizer

    Anthropic Agentic Systems - #5. Evaluator-Optimizer

    Anthropic Agentic Systems: A Five-Part Exploration is sponsored by Agent.ai - Discover, connect with and hire AI agents…

    1 条评论
  • Anthropic Agentic Systems - #4. Orchestrator-Workers

    Anthropic Agentic Systems - #4. Orchestrator-Workers

    Anthropic Agentic Systems: A Five-Part Exploration is sponsored by Agent.ai - Discover, connect with and hire AI agents…

    5 条评论
  • Anthropic Agentic Systems - #3. Parallelization

    Anthropic Agentic Systems - #3. Parallelization

    Anthropic Agentic Systems: A Five-Part Exploration is sponsored by Agent.ai - Discover, connect with and hire AI agents…

    1 条评论
  • Anthropic Agentic Systems - #2. Routing

    Anthropic Agentic Systems - #2. Routing

    Anthropic Agentic Systems: A Five-Part Exploration is sponsored by Agent.ai - Discover, connect with and hire AI agents…

  • Challenges with Operator

    Challenges with Operator

    Yesterday I had written about my initial response in working with OpenAI Operator. Here is a link to that article.

    4 条评论
  • Will Operator take my job

    Will Operator take my job

    There were 2 Tsunamis that hit the AI world last week - DeepSeek and OpenAI's Operator. This post is about Operator and…

    15 条评论

社区洞察