I think the overarching trend is "Buy v. Build" has reversed as hard as Roe v. Wade
If we haven't already crossed this point, the time that goes into software procurement, implementation, hand-off with the vendor, talking to support, getting customization will be less than just making something turnkey that solves exactly your problems
But we're definitely at the point already where building something quickly with AI is a already much more fun and rewarding use of time for any semi-technical person
I think this is true if you're the kind of "non-technical" person who reads HN which is still the minority
The smallest amount of framing and architectural forethought pays massive dividends but I imagine the person who says "build me an accounting app" while being apathetic to what language and stack it uses like apps such as Loveable imagine will still get bad results
In reality, VC is probably the most threatened because value and investment dollars have increasingly accrued to large public companies and a handful of a growth companies, the power law has never been stronger and returns never more stratified.
You can't just build a fund throwing out money at Seed and Series A SaaS companies anymore, more than ever company spend is going towards a few AI providers as "buy vs build" shifts in the opposite direction than before
And additionally you could argue that LLM coding is replacing the need for much of pre-Seed and Seed money that would go towards hiring the first 1-3 engineers and MVP development
I think the panic around distillation misses the fact that US labs also benefit heavily from Chinese breakthroughs like Deepseek's work on sparsity, MoE and training architecture
It may be that US labs use Chinese models for distillation but we'd ofc never know because they can host the models themselves
I'd implore anyone interested in metaprogramming to look at Lean4
It gets overshadowed by the theorem proving but it's unsurpassed in metaprogramming, to my knowledge it can do anything Lisp Racket and Rhombus can and much more
Even though Python code may use more characters/LoC than say Rust in text form, it's not necessarily more token dense because LLM tokenizers are good at "compressing" its English keywords
In contrast, langs with symbol-heavy syntax (ALP as extreme example) use fewer characters but don't tokenize well in practice so aren't as efficient as one would think
What does it bring to the table that isn't achievable with Cargo?
I just feel like Dart doesn't offer enough for a fairly standard OOP lang which isn't particularly fast and doesn't have the library ecosystem or vast training data of Rust, Go etc