Before GenAI I had a good sense of how long it would take me to build something… so i wouldn’t bother if it was a personal project and would take too long (to do right, to achieve feature levels, etc)
Now with AI… i have absolutely no idea, so I’m tackling personal projects and completing them
Maybe there will be a “free tier” Gen AI offering you heavily advertising biased content. Only the people with deepest pockets will be able to dial down the marketing bias delivery.
We aren’t seeing the true cost of LLMs. It is still subsidised by CapEx funding.
As OpEx starts to kick in, none of the GenAI players will be able to resist filling the vacuum left by inorganic search engine advertising.
The level of sophistication of future marketing is going to be off the charts. We haven’t even begun to scratch the surface.
There’s going to be a race to the next advertising game…
Ads in your generative AIs.
Let that sink in.
2030: Your coding assistant is going to fire ads at you.
Because you’re not using Google
They have to get you somehow.
Let’s say I’ve vibe coded a new product. Ok how do I sell it?
Throw it onto an App Store?
Let’s say it’s SaaS? How do I market it? Nobody is using search engines, so how do I reach my audience without cash? Suddenly the barrier to entry is both lower and higher in different places.
Also loads of people are going to find that writing the “perfect software” isn’t the challenge: it’s usually adoption! Nobody actually wants to buy perfect they want to buy what they’ve always bought. Adoption of innovation moves at the speed of humans not AI
Faster sales team, faster user research, faster product design, faster compliance… loads of bottlenecks around just writing code.
I also think the bottlenecks are elsewhere, as in not in the writing of software… so companies will find themselves scratching their heads trying to work out how to go about “faster”.
Until the environment around software engineering can go faster, what’s the point?
People keep saying AI is going to replace humans in dev teams… I think what’s more likely is that AI is going to increase the cost base of running a dev team… think every dev gets a $200 per month subscription plus additional credits… which means you can afford less humans overall.
I was being generous
Yes generative AI is shaking up software engineering right now, but there is growing evidence that techniques like (A)TDD and hexagonal architecture are helping teams adopt.
People that don’t understand what good “digital delivery” looks like might view “digital” as akin to throwaway code seen produced by digital marketing agencies… dismissing the good engineering techniques applied at scale and at pace across the public sector.
If you really want to slow down programmes of work around your system, assuming it has an API and lots of systems want to connect to it, make it really expensive to spin up new environments for testing. Make it cost the same as a whole production environment, and take multiple weeks/months of time
We keep swinging wildly as an industry between “everything should be local!” to “thin clients and mainframes!”…
Framing as a binary choice loses all nuance about why you might want to do both.
Apple seems to be serious about keeping development local and secure… but industry energy seems to be disproportionately spent on learning how to build out remote environments right now.
I think we need both. In parallel. They are different offers solving different problems.
It seems to be all the rage to move to remote environments… I think that they solve some problems but they aren’t a panacea. You can’t mitigate all threats by moving the execution environment into a virtual machine in a datacenter.
Plus there are situations where you need local dev.
Providing corporate laptops for people who write code and deploy software to production is nothing like providing laptops for people who don’t.
It is more similar to providing a platform to software teams. It requires deep understanding of the nature of their work and the way that dev tools work
Do you know what the best way to slow things down?
- do more stuff in parallel
- add more people
The winning formula always has been a) laser focus and b) small teams
We need to make public transportation cheaper in the UK. Yes, we could explore lowering fares but is this sustainable? What about making public transport (including to your usual place of work) tax deductible. Then sell volume train fares to employers to create stable train revenue demand.
There is no such thing as a “business strategy” nor “tech strategy”. There is just “the strategy”. People separate them but they always discover failure or compromise.
So long as I get the department of bins I’m happy
Recognising that we’ve always had people who have never figured it out…
- more tension and disagreements over constructive conversations
- less “overall engagement” in the work
- decreased empathy
At a time when “being human” (combined with our ability to reason) is our literal USP over AI.
Hypothesis here that a higher proportion of anyone entering the workplace post 2020 compared to pre (especially in knowledge work) are going to be super in the dark about what good looks like for a real long time (maybe forever).
In a world where you don’t regularly see your colleagues f2f… I think people forget how important interpersonal relationships are, and how difficult they are to maintain over video calls. It requires active effort from BOTH parties, even more tricky for groups to resolve interpersonal dynamics.
There are a narrow range of good answers to this question haha
I am also very sceptical of profiles that are overwhelmingly full of AI experience? Why are you submitting your CV for this mediocre day rate job if this is true? Does nobody want to work with you in well funded startup land? Are these claims on your CV true??
Because *everyone* is trying to experiment to release the first good product market fit / scalable and stable products.
*expertise* in this space is looking one layer down on the wardley map. Does the individual show expertise in projects in genesis, have they experimented a little with AI?
Hmm, the usual approach of “has >5 years real world experience in X” to determine if someone is good does not apply to cutting edge things like AI engineering. Things that are on the left of the wardley map… you can’t “buy” the knowledge without spending extreme amounts of money or getting lucky.
