[HN Gopher] Technical Deflation
___________________________________________________________________
Technical Deflation
Author : 0x79de
Score : 71 points
Date : 2025-11-24 12:39 UTC (3 days ago)
(HTM) web link (benanderson.work)
(TXT) w3m dump (benanderson.work)
| Joker_vD wrote:
| > when prices go _down_ instead of up. It is generally considered
| harmful: both because it is usually brought on by something
| really bad (like a severe economic contraction)
|
| Or, you know, technological improvements that increase efficiency
| of production, or bountiful harvests, or generally anything else
| that suddenly expands the supply at the current price level
| across the economy. Thankfully, we have mechanisms in place that
| keep the prices inflating even when those unlikely events happen.
| marcosdumay wrote:
| Deflation is about _all_ prices going down. Just a few
| decreasing is normal.
|
| Anyway, WTF, economics communication has a huge problem. I've
| seen the article's explanation repeated in plenty of places,
| it's completely wrong and borderline nonsense.
|
| The reason deflation is bad is not because it makes people
| postpone buying things. It's because some prices, like salaries
| or rent just refuse to go down. That causes rationing of those
| things.
| jdasdf wrote:
| >It's because some prices, like salaries or rent just refuse
| to go down.
|
| a common argument, but one that doesn't bear out in the
| absence of regulation enforcing that.
| gus_massa wrote:
| I agree. It's super common that the price of vegetables goes
| up and down arround the year, in particular due to the
| harvest season.
| HPsquared wrote:
| Rent and salaries don't like going down because of debt.
| Debts are denominated in currency units and go up with
| inflation (interest rates have a component to correct for
| inflation) but they don't decrease if the currency _gains_
| value over time (this would need negative interest rates). I
| suppose that 's something that could be done with regulation.
| marcosdumay wrote:
| Yes. And even if people could refinance, debt values going
| down causes further deflation.
| HPsquared wrote:
| Does it? Debt repayments are money deletion, so if debt
| is nominally written-off, less has to be paid back. That
| is, there will be less "anti-money" in the system but the
| "money" is still there. That increases the money supply,
| therefore inflationary.
| pjc50 wrote:
| See "price stickiness" and what is simplified as "menu
| reprinting costs"; there's usually a cost associated with
| changing prices, and a cost associated with _renegotiating_
| prices for everything that 's not being sold on a spot
| market. People cannot buy housing at spot, and while spot-
| labour pricing is definitely a thing for some services it's
| so socially destabilizing for anything skilled that most
| workforces operate on salary.
|
| The reverse of this is that high inflation tends to cause a
| lot of strikes, because salaries refuse to go _up_ and very
| high levels of inflation need salary repricing every month or
| even week.
| igleria wrote:
| In Argentina I've learned from a young age that prices take
| the elevator, but salaries take the stairs.
|
| It got old really quick having to negotiate with the boss
| every 6 months.
| darkerside wrote:
| Does anyone else agree with this the premise of this article? Is
| it sensible to put off building things now because it will get
| even cheaper and faster later?
|
| Maybe the time value of time is only increasing as we go.
| blueflow wrote:
| Actually yes. I wanted to get into UI programming with GTK 2
| and right now im waiting for GTK $n to stabilize so i can
| commit to it.
|
| Knowing that GTK $n-1 will soon be obsolete is enough reason to
| not put effort into learning it.
| lo_zamoyski wrote:
| In general, we should focus more on what endures over what
| changes. Focus less on the times and more on the eternities.
| People very often drown in the noise of passing fads and
| fashions and ephemeral tech. Can you become really skilled at
| using some piece of tech? Sure. Is it worth becoming _really_
| skilled? It depends on the circumstances and the particular
| person, but in most cases, probably not. It usually is a
| waste of time (but given the kinds of SFVs that people
| publish or hobbies people have, people are generally quite
| good at frittering away their lives on stupid shit).
|
| Incidentally, this is how you can distinguish between a good
| CS curriculum from a bad one. A good one focuses heavily on
| principles; the particular technical trappings are mostly
| just a medium, like Latin used to be in academia, now
| replaced by English. You pick up what you need to do to the
| job.
| hahajk wrote:
| The conclusion that you should wait to build anything is an
| illustration of the danger of economic inflation that the
| author started with. I'm not sure why he thinks the economic
| version is toxic but the technological version is a good idea
| though.
|
| The answer to should we just sit around and wait for better
| technology is obviously no. We gain a lot of knowledge by
| building with what we have; builders now inform where
| technology improves. (The front page has an article about
| Voyager being a light day away...)
|
| I think the more interesting question is what would happen if
| we induced some kind of 2% "technological inflation" - every
| year it gets harder to make anything. Would that push more orgs
| to build more things? Everyone pours everything they have into
| making products _now_ because their resources will go less far
| next year.
| philipallstar wrote:
| > I think the more interesting question is what would happen
| if we induced some kind of 2% "technological inflation" -
| every year it gets harder to make anything. Would that push
| more orgs to build more things? Everyone pours everything
| they have into making products now because their resources
| will go less far next year.
|
| Government bonds already do this for absolutely everything.
| If I can put my money in a guaranteed bond at X%/year then
| your startup that's a risky investment has to make much
| better returns to make it worth my while. That's why the
| stock market is always chasing growth.
| dvh wrote:
| I agree. I had several projects lined up and I delayed one
| because it used same tech as another significantly smaller
| project, so I learned the tech on the smaller simpler project
| and then used the knowledge on the bigger project. It was
| beneficial to not do the bigger project first.
| Ray20 wrote:
| > it will get even cheaper and faster later
|
| Yeah, and will be done by somebody else. I think this is the
| main problem, and if you get rid of it, you'll have a
| completely sensible strategy. I mean there are many government
| contractors who, through corrupt connections, can guarantee
| that work will be awarded to them, and very often doing just
| that.
| bootsmann wrote:
| I think you're right. The author is quite wrong on many aspects
| in my view. One of the central mistake he makes is that
| creating a profitable startup is mostly a matter of shipping
| good product i.e.
|
| > Used to be, you had to find a customer in SO much pain that
| they'd settle for a point solution to their most painful
| problem, while you slowly built the rest of the stuff. Now, you
| can still do that one thing really well, but you can also
| quickly build a bunch of the table stakes features really fast,
| making it more of a no-brainer to adopt your product.
| Zigurd wrote:
| Technology has always been deflationary. But you don't put off
| buying a computer because it will be cheaper next year. Nobody
| seems to be putting off buying GPUs despite scary depreciation
| and a blistering pace of new product introductions that are ever
| cheaper faster and better.
| mr_toad wrote:
| Some people do put off buying cellphones and laptops when they
| know a new model will come out every year.
| tikhonj wrote:
| The overall trend has been the opposite though, hasn't it?
| People used to buy a new phone (or new laptop/etc) every
| couple of years because the underlying tech was improving so
| quickly, but now that the improvements have slowed down,
| they're holding onto their devices for longer.
|
| There was an article[1] going around about that recently, and
| I'm sure there are more, but it's also a trend I've seen
| first-hand. (I don't particularly care for the article's
| framing, I'm just linking to it to illustrate the underlying
| data.)
|
| [1]: https://www.cnbc.com/2025/11/23/how-device-hoarding-by-
| ameri...
| aleph_minus_one wrote:
| > Some people do put off buying cellphones and laptops when
| they know a new model will come out every year.
|
| Don't confuse technical deflation with the Osborne effect:
|
| > https://en.wikipedia.org/wiki/Osborne_effect
| raw_anon_1111 wrote:
| The Osborne affect has been heavily disputed over the years
| - it says so in your own citation.
|
| But at least with iPhones, there is a deflationary affect
| because Apple has since the 3GS in 2009, kept the old phone
| around and reduced the price. For instance my son wanted an
| iPhone 16 Plus. I told him to wait until the 17 was
| announced and he bought one cheaper from T-Mobile
| vrighter wrote:
| only really faster and better if you _don 't_ use them for
| gaming, unfortunately. Upscaling and frame generation is not a
| better GPU, it's one with a band-aid applied to hide the fact
| that it actually did _not_ get much faster.
|
| RTX doesn't count to me either, because that's some bullcrap
| pushed by gpu manufacturers that requires the aformentioned
| upscaling and frame generation techniques to fake actually
| being anywhere close to what gpu manufacturers want gamers to
| believe.
| Aurornis wrote:
| > only really faster and better if you don't use them for
| gaming, unfortunately. Upscaling and frame generation is not
| a better GPU, it's one with a band-aid applied to hide the
| fact that it actually did not get much faster.
|
| The generations gains haven't been as great as past
| generations, but it's getting silly to claim that GPUs aren't
| getting faster for gaming.
|
| Intentionally ignoring frame generation and DLSS up scaling
| also feels petty. Using those features to get 150-200fps at
| 4K is actually a very amazing experience, even if the purists
| turn their noses up at it.
|
| The used GPU market is relatively good at calibrating for
| relative gaming performance. If new GPUs weren't actually
| faster then old GPUs wouldn't be depreciating much. Yet you
| can pick up 3000 series GPUs very cheaply right now (except
| maybe the 3090 which is prized for its large VRAM, though
| still cheap). Even 4000 series are getting cheap.
| vrighter wrote:
| "Guessing what a pixel's color might be were one to
| actually do the work and render it" is not the same as
| actually rendering it. No, upscaling doesn't count.
|
| Doing it for a whole screenful of pixels, for the
| _majority_ of frames (with multi-frame generation) is even
| less of it.
| toast0 wrote:
| I dunno, the kiddo went from a 1650 Super to a 3060 and it's
| a lot nicer looking, I don't think frame gen and what not is
| enabled. Sure, that's up a notch on the SKU list and tons
| more VRAM. The 1650 Super was working with most of the games
| he tried, but Marvel Rivals was terrible (haven't seen him
| play it with the new card though)
|
| It does help that he has a small screen and regular DPI.
| Seems like everyone wants to run with 4x the pixels in the
| same space, which needs about 4x the GPU.
| Ray20 wrote:
| > But you don't put off buying a computer because it will be
| cheaper next year.
|
| Why not? Sounds like a pretty reasonable strategy.
|
| > Nobody seems to be putting off buying GPUs
|
| Many people doing exactly that.
| detourdog wrote:
| Back when I was first started wrestling with this issue it
| the question was "how much faster will the daily photoshop
| operations be with a new computer?"
|
| Now a new computer barely does anything faster for me.
| Glemkloksdjf wrote:
| I'm still ahead of a company starting later:
|
| I have the legal structure, i know my collegues, i have
| potentially employees and more capacity.
|
| The problem is not that a startup is starting after you but you
| do not give yourself time to keep an eye on AI and not leveraging
| it when its helpful.
|
| We leverage AI and ML progress constantly and keep an eye on
| advances. Segment Anything? Yepp we use it. Claude? Yes Sir!
| jackar wrote:
| The impulse to make the comparison makes sense. The reality is a
| bit different, probably leans in the right direction but is
| buffered by learning. I'll explain. There is no upside to
| delaying purchasing if you think things will get less expensive.
| There is however upside in building today even if you have to
| rebuild tomorrow, and that upside is in learning the problem
| space. Specifically, what is likely to be trivialized and what
| truly requires domain knowledge. Horizontal apps? Little domain
| knowledge to encode. Vertical app? More domain knowledge to
| encode. Separately, there are more ways to differentiate than
| distribution alone, see verifier's law. Problems that are
| challenging to verify are challenging for AI to trivialize.
| readthenotes1 wrote:
| "One of the main problems is that if people expect prices to keep
| going down, they'll delay purchases and save more, because they
| expect that they'll be able to get the stuff for less later."
|
| That is why we are all waiting to buy our first personal
| computers and our first cell phones.
|
| Economists have managed to be ludicrous for a very long time and
| yet we still trust them.
| api wrote:
| There's a fun version of this in futurist space travel
| speculation.
|
| Let's say you have the a fusion rocket and can hit 5% the speed
| of light. You want to migrate to the stars for some reason.
|
| So do you build a generational ship now, which is possible, or...
| do you wait?
|
| Because if you build it now someone with a much better drive may
| just fly right past you at 20% the speed of light.
|
| In this one the answer is to plot it out under the assumption
| there is no totally undiscovered major physics that would allow,
| say, FTL, and plot the curves for advancement against that.
|
| So can we do this with software? We have the progress of
| hardware, which is somewhat deterministic, and we know something
| about the progress of software from stats we can make via GitHub.
|
| The software equivalent of someone discovering some "fantasy"
| physics and building a warp drive would be runaway self-improving
| AGI/ASI. I'd argue this is impossible for information theoretical
| reasons, but what if I'm wrong?
| sebastianconcpt wrote:
| Dang. I won't be able to sleep tonight
| raw_anon_1111 wrote:
| What I don't understand is that how is this automatically good
| for startups?
|
| Say I have a startup that vibe codes "AI for real estate". What
| about customer acquisition?
|
| On the other hand, if I'm Zillow, why can't I just throw a
| developer on the same feature and automatically have a customer
| base for it?
|
| If you look at most of the YC funded startups these days, they
| are just prompt engineers with no go to market strategy and some
| don't even have any technical people and are looking for
| "technical cofounders" that they can underpay with a promise of
| equity that will statistically be meaningless.
| Bombthecat wrote:
| Yap, AI coding will just help monopolize positions lol.
| raw_anon_1111 wrote:
| Clay Christenson in the "Innovators Dilemma" would call AI a
| "sustaining innovation" not a "disruptive innovation".
|
| At the end of the day. Open AI is losing billions of dollars
| and Google caught up while still seeing record revenues and
| profits using its own infrastructure and TPUs
|
| Even the laggard Apple is reportedly just going to throw a
| billion (chump change) at Google for its model and keep
| selling phones and other hardware while OpenAI is reportedly
| working on a "smart egg"
| skeeter2020 wrote:
| >> First, models getting better makes AI-based applications
| easier to build, because they can be simpler.
|
| Don't conflate easy with simple. I'd argue they are actually
| easier and far more complex.
| JacobiX wrote:
| In the end, the article says:
|
| > writing functioning application code has grown easier thanks to
| AI.
|
| > It's getting easier and easier for startups to do stuff.
|
| > Another answer might be to use the fact that software is
| becoming free and disposable to your advantage.
|
| For me, the logical conclusion here is: don't build a software
| startup!
| cootsnuck wrote:
| Yup. I'm starting to wonder if the startup space has a pretty
| big blind spot not realizing that how easy it is to build
| mostly/semi functioning software is not a unique advantage...
|
| I left an AI startup to do tech consulting. What do I do? Build
| custom AI systems for clients. (Specifically clients that
| decided against going with startups' solutions.) Sometimes I
| build it for them, but I prefer to work with their own devs to
| teach them how to build it.
|
| Fast forward 3+ years and we're going to see more everyday SMBs
| hiring a dev to just build them the stuff in-house that they
| were stuck paying vendors for. It won't happen everywhere.
| Painful enough problems and worthwhile enough solutions
| probably won't see much of a shift.
|
| But startups that think the market will lap up whatever they
| have to offer as long as it looks and sounds slick may be in
| for a rude surprise.
| raw_anon_1111 wrote:
| Of course it still makes sense to have a startup. Not because
| you will ever find a decent enough market. But if you are
| well connected enough you can find a VC and play with other
| people's money for awhile.
|
| You aren't doing it to get customers, it's for investors and
| maybe a decent acquisition
| MangoToupe wrote:
| > Fast forward 3+ years and we're going to see more everyday
| SMBs hiring a dev to just build them the stuff in-house
|
| I don't see this happening. Businesses generally want
| familiar tools that work reliably with predictable support
| patterns.
| sgt101 wrote:
| The models themselves represent the biggest deflation cases I've
| ever seem.
|
| The charged cost of a frontier model is ~200x lower than 2 years
| ago, and the ones we are using now are much better - although
| measuring that and how much is challenging. Building a "better
| than GPT-4" model is also vastly cheaper than building GPT-4
| was... perhaps 1/100th?
| kittikitti wrote:
| The is a great point that wasn't included in the original
| article. Thank you.
| einpoklum wrote:
| > _Desktop app.... though Electron and Tauri have made it easier_
|
| Ugh. I don't like that kind of 'desktop' apps. Huge bloat with a
| blip of actual app.
| vrighter wrote:
| "But building the same functionality has undoubtedly become
| simpler."
|
| I disagree with this statement. It has become simpler, provided
| you don't care about it actually being correct, and you don't
| care about whether you really have tests that test what you think
| you asked for, you don't care about security, and other things.
|
| Building the same thing involves doing the things that LLMs have
| proved time and again that they cannot do. But instead of writing
| it properly in the first place, you now need to look for the
| needle in the haystack that is the subtle bug that _invariable_
| get inserted by llms every single time I tried to use them. Which
| requires you to deeply understand the code anyway. Which you
| would have gotten automatically (and easier) if you were the one
| writing the code in the first place. developing the same thing
| _at the same level of quality_ is harder with an LLM.
|
| And the "table stakes" stuff is exactly the thing I would not
| trust an LLM with for sure, because the risk of getting it wrong
| could potentially be fatal (to the company, not the dev. Depends
| on his boss' temperament) with those.
| raw_anon_1111 wrote:
| And subtle bugs don't get inserted by humans? Did security
| flaws in software just start happening after LLMs were
| introduced?
| vrighter wrote:
| So? The point is that humans do it _much_ less often.
|
| Let's say there are 10 subtasks that need to be done.
|
| Let's say a human has 99% chance of getting each one of them
| right, by doing the proper testing etc. And let's say that
| the AI has a 95% chance of getting it right (being very
| generous here).
|
| 0.99^10 = a 90% chance of the human getting it to work
| properly. 0.95^10 = only a 60% chance. Almost a coin toss.
|
| Even with 98% success rate, the compounding success rate
| still goes down to about 81%.
|
| The thing is that LLM's aren't just "a little bit" worse than
| humans. In comparison they're cavemen.
| raw_anon_1111 wrote:
| So humans do it much less often yet we have 30 years of
| evidence to the contrary? Humans still can't figure out how
| to write code not subject to sql injection after 25 years
| or how to write code and commit it to GitHub without
| exposing admin credentials
| peab wrote:
| I think either you haven't used LLMs for coding in a while, or
| you're working on things where they might still be limited on.
|
| I've been able to use LLMs to build things in a weekend that I
| would not have been able to do in the past, without putting in
| months of serious effort.
|
| I recently rewrote from scratch in a weekend a project that i
| had made a couple years ago. In a single weekend i now have a
| better product than I did at the time, when I spent maybe 20x
| the amount of time on it.
| moralestapia wrote:
| With re. to the first two paragraphs, it's crazy how someone can
| be so massively brainwashed.
| Sevii wrote:
| Good explanation for why hiring stopped. If AI is improving
| rapidly why hire engineers now that you might not need in 6-12
| months?
| locknitpicker wrote:
| > Good explanation for why hiring stopped. If AI is improving
| rapidly why hire engineers now that you might not need in 6-12
| months?
|
| I'm not so sure that's the reason. I mean, to believe LLMs
| replace engineers you first need to believe engineers spend the
| bulk of their time typing frantically churning out code in
| greenfield projects. That's not compatible with reality.
| Although LLMs excel at generating new code from scratch, that
| scenario is the exception. Introducing significant changes to
| existing projects still requires long iterations which
| ultimately end up consuming more development time than actually
| rolling out the changes yourself.
|
| The truth if the matter is that we are not observing an
| economic boom. The US is either stagnant or in a recession, and
| LLMs are not their cause. In an economic downturn you don't see
| spikes in demand for skilled workers.
| eldavido wrote:
| Hard agree with both points--this feels way closer to reality
| than most of what I've read.
|
| On recession: cost of living is becoming crisis-level. I read
| recently that 67% of Americans are paycheck-to-paycheck.
| 150k/yr is 12k/month. If groceries go from 500 to 1000/month,
| a 150k wage-earner save less for retirement. For someone
| making 30-40k (basically minimum wage), it's a huge hit. Then
| consider it's the same story for cars, housing, medical
| care...it goes on and on. It doesn't look "recessionary"
| because GDP keeps going up. But we're getting so much less
| for it with every passing year.
|
| I also agree that we need to consider what brownfield dev
| looks like. It's where the vast majority of my time has gone
| over 15+ years in software and I'm not convinced all the
| coordination / sequencing / thinking will be assisted with
| LLMs. Particularly because they aren't trained on large
| proprietary codebases.
|
| What we might both be missing, is that for most people,
| writing the actual code is _hard_. LLMs help with that a lot.
| That 's what a lot of junior/entry-level work, actually is
| (not as much planning/thinking as seniors do).
| Bombthecat wrote:
| We don't hire juniors anymore
|
| They are definitely not needed anymore.
|
| The market is flooded with seniors.
|
| So no problem there either
| zkmon wrote:
| Infact, the opposite is true. There is tech inflation. Coding
| will get easier, but that doesn't mean tech is getting easier. It
| will only get more complex and will drive more segregation.
|
| Tech is dividing the society and driving a wedge deeper. There is
| a huge population that are being thrown wayside by the high-speed
| tech highway. Which means the tech is getting more and more
| unreachable.
|
| AI assistants are only going to make this worse, by removing the
| direct touch between users and the tech. Tech becomes just
| unmanageable for average person.
|
| Just like how you were able to do all repairs for your bike, as a
| kid. But you can't do the same for your car now. Tech never gets
| easier or reachable.
| coolThingsFirst wrote:
| I think this is a very solid take. Moreoever, tech will be
| optimized for LLMs and not humans. Shitty languages will remain
| shitty because there's no need to optimise them or make them
| more elegant. Web dev will remain ultra fragmented and will get
| even worse as the years go by.
|
| Not going to lie, it looks like a bleak future.
| lubujackson wrote:
| What will become apparent is that when coding costs go to 0,
| support and robustness costs will be the new "engineering"
| disciple. Which is in reality how things work already. It is why
| you can have open source code and companies built on providing
| enterprise support for that code to companies.
|
| If you want to build a successful AI company, assume the product
| part is easy. Build the support network: guarantee uptime, fast
| responses, direct human support. These are the shovels
| desperately needed during the AI gold rush.
| makapuf wrote:
| Did google search and stackoverlow and internet deflate
| programming ?
|
| Comparing with reading books you might have or need to order,
| read, or get from the library, _you bet_.
|
| There still are some interesting problems to tackle. Maybe more
| than before. So who knows.
| robotburrito wrote:
| Does anyone really read any of this stuff anyway?
| all2 wrote:
| I take great pride in my role as a user of HN. This includes
| never reading the article and always positing very opinionated
| tasks without qualification.
| kittikitti wrote:
| One indicator that I'm watching for deflation is the concept that
| money is not accepted. Things like social media followers or
| likes are used as substitutes. People can't buy things like
| concert tickets or latest merchandise because all the bots have
| already bought them. Businesses will claim they can't provide a
| custom service. In tech, this would be spending a modest amount
| of money on online ads when in reality an ad campaign can only be
| "bought" with debt.
|
| I think analysis around inflation, deflation, and consumer prices
| are valid but they are part of an understanding from economies of
| 100 years ago. Money loses value when you can't do anything with
| it. Tech and AI runs on debt, and an extraordinary amount of it.
| Is that really money? I don't think so.
|
| Deflation may suffer from Goodhart's law. Because we've
| repurposed all of available human resources for mitigating
| against it, the variables we used to measure it cease to become
| useful. Our central measure for the economy are things like the
| stock market and the unemployment rate which have prevalent and
| valid criticisms that policy makers ignore. They truly don't
| indicate what's occurring on main street and I'm afraid that we
| will be in a deflationary spiral without knowing it.
| fasbiner wrote:
| I disagreed with some points and agreed with others based on my
| experience and the data I have available, but the last few
| sentences really weakened the overall point.
|
| > Giga AI, a company building AI customer support agents, claims
| to have sworn off the "forward deployed engineer" model of custom
| software favored by many other successful startups, in favor of
| software that customizes itself--only possible because of coding
| agents.
|
| Giga AI is not a publicly traded company and they have zero legal
| liability or possible downside for lying, and massive upside for
| lying. They also don't have real customers and are not in
| positive revenue. The trend is that everyone who has said this
| was lying.
|
| When there's tangible evidence of this, I think it will be an
| important part of the discussion. Until then, saying "claims" and
| "but I don't really know" but then paraphrasing their press
| release without analysis is about as sophisticated and as honest
| as tweeting "people are saying."
|
| The author should take their own advice and wait six months when
| these claims will be easier to substantiate and support the
| analysis far more strongly.
| hollow-moe wrote:
| Imagine, people aren't spending already because they have no
| money, things keeps getting pricier and pricier, people buy less
| and less trying to put aside in case something happens like an
| accident. And "higher people" still thinks "lower price bad". I
| sure hope the wall hits so hard countries fall.
| Barathkanna wrote:
| As someone building AI infrastructure, this deflation idea shows
| up very differently for us. Application founders can afford to
| say they will wait for the next model, but infrastructure
| founders cannot. The value we create is not in shipping features
| faster. It is in absorbing the volatility that sits underneath
| the entire AI stack.
|
| Models keep getting cheaper and more capable every few months.
| However, the underlying compute economics do not deflate at the
| same rate. GPU provisioning, inference orchestration, bandwidth
| constraints, latency guarantees, regulatory requirements, and
| failure handling do not become magically simple because a new
| model improved its reasoning. In reality, each improvement on the
| model side increases pressure on the infrastructure side. Bigger
| context windows, heavier memory footprints, more parallel
| requests, and more complex agentic workflows all increase the
| operational burden.
|
| For infrastructure teams, waiting does not help. The surface area
| of what needs to be built only grows. You cannot delay
| autoscaling, observability, scheduling, routing, or privacy
| guarantees. Applications will always demand more from the
| infrastructure, and they will expect it to feel like a commodity.
|
| My view is that technical deflation applies much more to
| application startups than to infrastructure startups. App
| founders can benefit from waiting. Infra founders have to build
| now because every model improvement instantly becomes a new
| expectation that the infra must support. The baseline keeps
| rising.
|
| The real moat in the next era is not the speed of feature
| development. It is the ability of your infrastructure to absorb
| the increasing chaos of more capable models while keeping the
| experience simple and predictable for the user
___________________________________________________________________
(page generated 2025-11-27 23:01 UTC)