Feels like the vibe coding hype made a lot of businesses confuse cheaper code with cheaper software development, нes, AI makes writing code faster. But software cost does not magically disappear because code gets generated faster.
It quietly moves:
more QA,
more review,
more debugging,
more rework later
And the irony is that many companies already started pricing AI engineering as if those costs no longer exist.
Feels like the discussion is slowly shifting from “AI will make engineering cheaper” to a more practical question: how do businesses adapt to what AI development actually costs?
Updated https://t.co/GMZpWFsPkb: more domain zones, faster models, search history, favourites, and a bunch of cleanup under the hood.
If you try it and think something is missing, or have ideas worth building - feel free to share
After a year of vibe coding hype, it increasingly feels like teams are generating more code than they actually understand.
I keep seeing the same pattern: something ships fast. Everything looks fine. But Nobody is fully sure what AI generated, what was changed manually later, or where the actual problem even starts.
The part people still underestimate is pretty simple: there is a big difference between using AI to write code and using AI to think for you
Feels like the AI era may need a new engineering culture: move fast if you want, but still understand what the hell you are shipping
Anyone surprised that the U.S. is turning AI regulation into a bit of a circus? I’m not.
While everyone watches the show around frontier model testing and changing narratives every few months, China is quietly becoming the open-source AI leader and building the infrastructure layer around its own technologies.
And no, China regulates heavily too. The difference is that while the U.S. debates control and monetization, China focused on regulating and building at the same time: they already require model filings, security reviews, and approvals for public AI systems.
But unlike the U.S. debate, the approach seems simpler: regulate the market, keep control, and still keep growth.
Ironically, they are doing it using the same playbook America historically used best: become the default stack everyone else quietly builds on and the question is how fast China take part of the AI market from the U.S.?
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
Reality is probably less dramatic than “one model changed everything.” Cyber capabilities may have accelerated the timeline, but this shift was inevitable the moment AI stopped looking like software and started looking like infrastructure.
Governments are usually anti-regulation right until a technology starts touching cyber, finance, or critical systems. Then suddenly “move fast” becomes “maybe we should have a framework.”
The funny part is that AI policy suddenly becomes “urgent” right around the moment frontier models start looking less like software and more like infrastructure.
For years the incentive was: move fast, scale, compete.
Now everyone is surprised that systems with cyber capabilities, financial impact, and geopolitical implications may need governance.
The real risk is not underregulation or overregulation. It’s building the rules only after a handful of companies become too critical to regulate without breaking everything.
I think the interesting shift here is that AI is already starting to move from “software/tool” territory into infrastructure territory.
Once a technology becomes capable of impacting cyber, critical systems, financial infrastructure, or large-scale automation, governments stop treating it like a normal product release.
The difficult part is that the same regulation meant to reduce systemic risk can also concentrate power around the few companies able to handle compliance at scale.
totally agree, we already see teams shipping tons of AI-generated code that nobody really understands anymore.
And the weird part is: most of it kinda works. Until it suddenly doesn’t.
A lot of AI code feels permanently “almost done”:
passes basic tests,
looks fine in demos,
but architecture, edge cases, and long-term maintainability are barely there
Crypto spent 15 years being called a solution without a problem. Turns out, the problem just hadn't arrived yet.
Now it's clear what it was built for - programmable money that moves without intermediaries, without paperwork, without a human in the loop. Not for memecoins or retail investors checking prices at 2am. For machines, and AI agents are exactly that.
They need wallets, not bank accounts, smart contracts instead of legal agreements, and onchain settlement instead of 3-5 business days. Everything crypto was criticized for being "too complex" or "too trustless" turns out to be exactly what autonomous systems need to transact at scale.
So what happens to the narrative when the actual user finally shows up?
@coinbase launched a marketplace where AI agents buy services from each other. No humans, no banks, no API keys.
Got curious what AI agent activity actually looks like across the whole industry — went digging.
167M transactions, $24M in monthly payments, 195% growth in one quarter. Sounds like the revolution already happened. Strip the noise and real economic activity is $1.6M.
Here's how it look: infrastructure exists, agents are already paying each other, but the economy isn't there yet. And with hype outrunning reality by 15x
🚨 Exclusive #ThoughtLeadership@dyadkov explains that the next phase of crypto growth may be driven by AI agents—not retail investors.
In previous #cycles, adoption came from ICOs, #DeFi, and NFTs, bringing in users through hype and narratives.
But the next wave could be different, autonomous AI systems using crypto to operate and transact.
“AI enables decision-making, while crypto enables value transfer,” highlighting how both technologies naturally work together.
As AI agents handle tasks like payments, data access, and compute usage, they could create real, utility-driven demand for crypto instead of speculation.
💡 This means the next crypto cycle may be powered by machine activity, not just human interest.
👉 Follow for more #crypto #Web3 updates
🔗 Full op-ed in comments
The tech and digital products market is entering a more mature era. 🌐
We spoke with Bitmedia Labs Founder Matvii Diadkov about building scalable systems and long-term success. 💡
Read more: https://t.co/eJfyJTWZgA
#AcquisitionInternational#BusinessExcellence#Technology