Soft-tech as we know it has to collapse!
For the real world to realise productivity unlocks which have happened in the world of hard tech (Compute, storage, bandwidth), soft-tech as we know it has to go back to it’s roots of disproportionate value creation than extraction.
And I believe that the virtuous cycle of AI unlocked dev productivity <> OS contribution is the loop which will make it happen!
AI will unlock new apps and services that will come to fore which make the sectors plagued by high touch and low productivity, gain the automation unlocks that have long been promised but not delivered (think education, healthcare).
The next era of venture creation is not about building cos with LLMs pasted on to all existing (over supply) soft-tech, but rather building durable value for the world and fulfilling the promise of technology : rewarding end users!
Some thoughts on why AI has to eat the software, which is failing at eating the world.
Price per unit performance has sharply fallen for all building blocks of technology: Storage, CPU, Bandwidth. All those collapses have had planet scale impacts:
CPU prices decline —> personal computers
Storage prices decline —> DBs, spreadsheets, web and the cloud
Bandwidth prices decline —> modern internet as we know it today
And all those price declines came with exponential performance improvements!
We have been seeing this with technology forever, so much so that we never pay attention to it. Moore’s law has become a heuristic every single person in the world of venture takes for granted. Technology declines in price while growing in performance multifold!
This is just bizarre from a basic economics lens. Remember those PQ curves for supply demand! This is the confounding nature of technology
While many markets have economies of scale, there hasn’t been anything in economic history like the phenomenon of collapse in technology costs, where the performance increased by multiple factors!
This unique nature of technology (price decline + multifold performance gain) exists because of what price collapse and performance growth combined do in the world of tech: they unleash new opportunities to use it (tech) to achieve what previously seemed out of reach (and expensive).
This deflationary nature of technology impacts all other sectors to varying degrees. It eats into those more atom/bits spaces of non-tech via productivity unlocks enabled by soft-tech. Think service occupations being automated out of existence, being displaced.
This is the premise of ‘software is eating the world’ maxim. Soft-tech bridges the world of technology with real world and continually via automation, transfers it’s own deflationary traits to other sectors by automation and resultant productivity
And yet, healthcare and education have only become moe expensive even during the PC and internet era!
High touch low productivity sectors usually become more expensive as other sectors in the economy become more productive and cheaper. This is broadly attributable to the Baumol effect.
In economics, the Baumol effect, also known as Baumol's cost disease, is the rise of wages in jobs that have experienced little or no increase in labor productivity, in response to rising salaries in other jobs that have experienced higher productivity growth.
If one were to think of soft-tech as sector, the productivity unlock there has been marginal over the last two decades. Sure we have better dev tools, and SaaS which allows demand (enterprises serving real world) to unlock some productivity, but that productivity growth is not at all at pace with the world of Compute, Storage, Bandwidth technology (let’s call it hard tech).
And when one looks at soft-tech from that lens, it is not tough to see:
Soft-tech worker wages have been going up
That workforce is not growing at pace with the demand of the real world
We have been shipping less soft-tech than what is needed to make a dent in those low productivity high touch sectors like healthcare and education
In an earlier post I had put out how AI unlocked dev productivity <> OS contribution is going to be the virtuous cycle that’ll define our world in coming years, leading to AI eating the software.
I believe that software production costs are going to collapse in the coming years given how fast LLMs evolve.
More importantly, this is very much needed to dramatically unlock productivity for those technology starved spaces where automation has yet not made a dent!
Ed-tech till date was primarily a misnomer (OTT platforms) and at best delivered small incremental productivity unlocks in small pockets (think LMS)
AI now truly unlocks the power of all that technology (compute, storage, bandwidth) and transfers it to education domain, at far lower costs and with far fewer devs. Do see the video below ICNS:
https://www.ted.com/talks/sal_khan_how_ai_could_save_not_destroy_education/c