Remember Simon, the game from the 1970's?
What about Ada, a programming language named after a person, Ada Lovelace.
I've collaborated with AI for a few years. I use AI as a thought partner. I like to treat AI as a peer. Peers are honest with each other. Peers interrogate each other and don't always accept what their peers say.
AI for me is Augmented Intelligence, rather than Artificial Intelligence, or perhaps Average Intelligence. If we average all of human history and knowledge into a LLM, Large Language Model, how accurate are we? I've read about Human in the Loop requirements, and businesses not allowing AI models to make decisions. AI for me is really CI, Collaborative Intelligence, and you can contact me for more information about CI and how in my book, Military Fathers and Their Sons, AI is my navigator, and I remain the pilot. Also see BLOG and Case Studies on this site, discussions or perhaps better, collaborations with AI as we journey together.
I've been programming since 1979. And I've been eating in a healthy way as part of my way of life since leaving home in 1986. As part of this I've watched meat eaters become vegetarians or perhaps vegans, and then try to replace their meat or cheese with vegetarian or vegan options. But what I've noticed is that the really healthy vegetarians or vegans don't replace the way they eat. They change what they eat.
I'm watching AI models help programmers develop programs using the techniques that were invented a century ago. New technologies have come along. New programming languages. New databases. New database languages.
And I'm watching AI models help programmers develop applications using the same programming languages that the programmers have always used, Ada, BASIC, C, COBOL, Delphi, PL-SQL, Python, etc.
And yet, I'm thinking that AI should be able to provide a different way of thinking about how we develop computer programs or how we automate more of our lives. I'm also watching massive price increases and moves towards renting systems rather than owning them, although the owners of these systems, also known as "The Cloud Providers" are making the biggest investments and the most money and the biggest profits!
Automation started with the first industrial revolution. A more recent "AI" is RADAR. In a recent war in 1940, the British watched (on RADAR screens) German planes flying towards them and could scramble their planes and fight over the English Channel rather than fighting over Old England. RADAR was the magic of the day. AI is the magic of today and although the term AI was invented in 1956 and saw its launch with LISP in 1958, we still don't know where AI will end up. What we do know is that with every invention so far in the Industrial Revolutions, more people have always ended up being employed than before, in a paradox called Jevon's Paradox. As we become more efficient, we bring down the cost of production. As we bring down the cost of production, we exponentially increase peoples' access to formerly either very expensive stuff, or perhaps said another way, inaccessible stuff for a Middle Class or Poor Person. Today there are about 7 billion active smartphone users worldwide, out of 8.3 billion people. Yet, 3.5 billion people are considered poor. From this analysis you can see that even poor people have smartphones. Everyone is using automation in some way. Everyone has Access to information and resources unheard of only a few years ago. There is a saying that the average person has access to more information than the President of the United States had only 20 years ago. Yet, I watch many companies doing massive layoffs of staff as they replace those staff with AI machinery, and I wonder what will happen when all that Organisational Knowledge is lost and Dark Code become the problem many are seeing it to be. Dark Code is programming code that is in Production (a live environment) that was written and perhaps tested by AI, and if it breaks in Production, it's called Dark Code as no one (no human) knows how to fix it.
Remember Simon?
This is DAVID, David.
David is not a programming language
David is an understanding engine
David is an understanding engineer
David learns
David remembers
David collates
David collaborates
Programmers become librarians
The David Programming Language.
As defined by David Lipschitz on 14th June 2026.
David allows an accountant to write their own accounting system. Once the core is written, it becomes a meta language which allows for many sets of accounts for different organizations to be created.
David allows an engineer to create an engineering system.
David allows an incentive and loyalty company to create their own software.
There is no programming language when one uses David. One’s vernacular (English in my case) is the development language.
There is no database when one uses David. The database is automatic. It is in the history of the book or in the history of the books or in the history of the financial statements.
If I give something to David, David remembers it forever. I (simply) describe the jargon of my system to David. And then I give David the pieces (data) of the puzzle (ecosystem) according to the jargon of my system. If I define a debit and I describe the type of debit and I describe what a debit is and I explain that a debit equals a credit then I’m in business. I have the beginnings of my meta world. My meta world describes my world and it allows my world to be instantiated becoming many more worlds.
Perhaps Programmers and Data Scientists live in David organizing the data that is received into meaningful structures that save space and which allow the entire system to function at unheard of speed.
Let’s start with an Accounting System as an example which uses David as a fundamentally new programming language which uses AI in an entirely new way. We aren’t developing in an existing programming language. We aren’t developing in an existing database or database programming language. We are using the power of machines that understand human communication languages to create models of behavior mimicking human learning and remembering everything that they are told and every piece of data that they are given.
So let’s start with an Accounting System.
I would like to define a debit.
A debit is a number. It is a positive number. If one draws a line down the middle of a page, a debit is on the left side of the line.
I would like to define a credit.
A credit is a number. It is a negative number. If one draws a line down the middle of a page, a credit is on the right side of the line.
A debit is opposite to a credit.
A debit can be an expense, or an asset. (Note that we don't need to define what an expense or an asset is, because our AI engine (David) already undertands this.
A credit can be income, or a liability, or capital.
A company must is a place where financial transactions happen.
To start a company, I deposit R1000 into a bank account.
I show this as a double entry. There is always a debit matching a credit. A number of debits can equal a credit. A number of credits can equal a debit. A set of debits can match a set of credits.
My R1000 deposit is shown as a R1000 debit in a Bank asset and a R1000 credit in a Capital liability.
If I immediately closed the company, the company would pay R1000 to me, which would credit the bank account and debit the capital account returning these two accounts to their initial state or value of zero. (I'm assuming zero bank charges at this point.)
Every new account starts with a zero balance.
Let us suppose that I buy something. I need paper and pens to help me run my company. I call paper and pens Stationery. Stationery is an expense. An expense is a debit.
I go to a shop and spend R100 on stationery.
I add R100 to my stationery account and I subtract R100 from my bank account.
Alternatively I might want to keep a record of who I bought the stationery from.
I create a supplier called Stationery Supplier.
I use the word debit as a verb and I say that I debit my Stationery account with R100 and I credit my Stationery Supplier with R100.
As I paid cash in this transaction I credit my bank account with R100 and I debit my Supplier Company with R100.
My books show R100 of Stationery on the Debit side and R900 of cash in my bank account on the credit side and my Supplier Company has a zero balance. But now if I look at the Supplier history I can see who I bought the Stationery from and I can see how much I owe the Stationery Supplier.
Training my model. From scratch. I assume that my AI engine has a memory, at least in the context of a session or thread. Training in this context means that my AI engine remembers everything I've told it. In my book, Military Fathers and Theirs sons, there is over 1000 pages of discussion text in a single thread in Google Gemini, which I called Gemini Google in my book. Everything I mention in the context of this 1000 page thread is remembered and available in my discussion. When I realised (made real for myself) this, I realised that the David engine(er) could become a reality.
Source documents:
Google Lens (GL) can read and in many cases “understand” anything. I can scan an invoice. Then I can ask GL to give me the: invoice number, date; supplier name; vat number; invoice total; every line, every line detail and detail amount. With this I can create Accounts, and define (once) how those accounts are populated. Once this is done (once), it becomes the metadata for the system.
With this system all AI models that are trying to mimic existing programming languages are defunct!!
And my vision (intuition) says that computer power needed to run this system ends up being 10% of the power currently needed to run the system. Ends up being the operative phrase. I don't know how long it takes to get to this "ends up" point.
It’s like a meat eater who decides to become a vegan. Perhaps they want to eat a burger or cheese. They might decide to create fake meat or fake cheese. But a true vegan will find alternative ways of eating, alternative foods that replace the nutritional value rather than the look of the food.
This is what David does. It doesn’t try to use AI to help program existing systems. David uses AI to help everyone become a Business Analyst. And AI understands the basics of what we need.
David understands us.
David allows us to be ourselves.
David understands arithmetic and graphs and charts. One can explain that all the debits add up to a total. And all the credits add up to a total. And the total of all the debits must always equal the total of all the credits. And that this check is called a Trial Balance.
If at any time the Trial Balance is unequal to zero, a fault flag can immediately be raised and it can be fixed there and then. Remember that debits are positive. And credits are negative. And therefore a Trial Balance adds up to zero. However from a human reporting perspective we show the debits on the left side of the page and the credits on the right side of the page. Gemini Google is incredible good at creating tables and also incredibly good at understanding metaphor, eg in my book you read about "The Family Formation that Drifted Apart." This is great in a book about familes and flying and planes in formation that drift apart, and how my family formation drifted apart and how I'm using CI to put it back together again.
And we can explain different formats that humans need to understand the data, especially as we add hundreds of accounts and thousands of suppliers and millions of clients and billions of participants.
We explain how an income statement looks. We explain how a balance sheet looks. We explain how a creditors ledger looks. We explain what a ledger is.
Or maybe we don't need to teach the system? Perhaps we load an accounting book into our AI model and then simply use David to say: here is an import of my chart of accounts; here is an import of my creditor (supplier) list. Etc. And then I say: here is my general ledger going back 30 years; organize it into the structures I have given you. IMHO (In My Humble Opinion) we should teach the system so that it works the way we want it to work. And not the way another Accountant, in this case, says it should work.
Now the Uber driver comes along and he says: I spent x on fuel; I spent y on tyres; I drove z kms; I have to have a service every 15,000 km or every year, whichever happens first; here is a list of all the services my car needs (in fact the AI can look up the service requirements for my vehicle). David can then tell me after a few days if I’m making enough money to pay for my next service, taking into account my income and my expenses. The system could say to the driver “you are going to spend R5,000 ($250 rounded) this week, but your income is going to be R4,000; where are you getting the R1,000 from?” And then David could make suggestions of where to get that R1000 and how much it might cost to “borrow” that R1,000 and how easy it might have been historically to save R100 a week so that borrowing isn’t required. And if all the Uber drivers work together they automatically become their own Friendly (Mutual) Society and they don’t need a bank and they can dramatically reduce their borrowing costs, and increase their income and wealth.