Greg Munt

What dont you like about waterfall and does it apply at AI speed?

This second fork will be whether Americans lie down and accept fascism, or they decide to fight. And I'm not talking about hand-wringing and letters of complaint.

If he fails, house and senate will go blue: perhaps by a wide margin. Trump will be out soon after.

If he succeeds and the official result is a 90%+ vote for red, there will be another fork in the road.

There are two paths ahead. Trump will accomplish a fascist takeover of the voting, or he will not. And obviously he is dumb enough to go for an overwhelming victory instead of a 49/51 split favouring red.

Cut out the execs and investors entirely. Developers need to be bootstrapping new businesses, right now.

That word is so overloaded with toxicity now. Say "agile" and people will react negatively. Describe agile (without saying The Word) and they are all over it.

His process appears to be:

- Use standard tools (linting, static analysis, security scanning, etc)

- Use LLM agents

- Human review only of things impossible or inapproopriate to automate

bdfinst.github.io/ai-patterns/...

This does not appear aligned with what you suggest?

Unless I have misunderstood, that he has everything automated - including code review.

Does this mean that you reject Bryan Finster's claims?

Is that because AI is fundamentally the wrong tool, or that people are trying to attach AI to pre-AI processes? Or that these pre-AI processes were dysfunctional all along and humans were mitigating them?

Isn't the premise that demand has fallen because of AI?

Despite Jakob's Law, I always railed against everything looking like everything else.

It hurts me a little, how passionately Jakob Nielsen has embraced AI.

I dont understand. I thought the whole point of this post is that it was nothing to do with AI?

Not saying I disagree. But... So what? Economies of certain countries completely obliterated because they have become dependent on AI companies?

In 2019, around 65000 layoffs. In 2025, various sources put it between double and quadruple that number.

The issue is not that there are less vacancies than 2021. It's that supply overwhelms demand.

There was a time when putting my CV out would result in 5-10 calls, the same day. Now? Nothing.

There was a time when an 80% skills match would have been more than enough. Now? Minimum 100%.

Local LLMs are the ONLY feasible future. Purely from a financial standpoint. When the reckoning comes, some BIG companies are going to BURN.

It's a "Why are you asking?" question.

Which I guess might be considered odd for an LLM... I personally would be curious rather than amused, though.

Why?

Seems like a reasonable response to me. Is AI autistic?

12 weeks seems generous.

Could it be that nobody has thought to, or is prepared to, install the high fences that are required? I'm only seeing the people who have put a LOT of effort into guardrails that are strong advocates.

For individuals, perhaps. For a team, often the opposite occurs.

They want reporting to light up this hole. When someone higher than them wants to know what the team is doing, a shoulder shrug is not going to go down well. Shrinking the hole (by delivering more frequently) will help way more than story points.

In my experience, the requirement for time estimates is dysfunctional or an indicator of dysfunction. Typically, delivery is taking too long. The period between deliveries is a black hole for management.

Instead of obfuscation, deal with the problem. Ask why management wants time estimates. Treat them as part of your team; because they are.

Use of story points is shifting or ignoring the problem. That problem: management wants time estimates and will convert them into deadlines.

My definition of "vibe coding" may differ from yours. For me, it's letting AI do what it likes without any kind of review at all. I would be shocked if <your favourite programming thought leader> was doing that.

Like "Never Forget", you mean?

Companies expecting their staff to just "do it", that's nothing new. See also: testing, UX, performance, security. Because training takes time, effort and money.

"AI is supposed to *save* us money, not be an additional training cost! How hard can it be?!"