- Author
- José Valim (Dashbit Blog)
- Published
- 24 September 2026
- Kind
- Company blog post of short opinion essays, in two parts
- Original
- dashbit.co/blog/evolving-ai-era
- Subject
- Coding with agents
What it is
José Valim writes on the blog of Dashbit. The post calls itself a digest of ramblings that may change. The page gives no other biography.
What it says
- Valim says agents may narrow gaps between ecosystems. Cheap code may also weaken the pull to share libraries.
- Valim says syntax conveniences matter less to agents. Valim calls languages built for agents around syntax short-sighted.
- Valim does not expect agents to replace compilers. An architecture-independent layer would reappear.
- Valim suggests languages favour stronger guarantees, such as explicit types, over inference that helps humans.
- Valim proposes a queryable program database of symbols and calls, instead of editor-style language servers.
- Valim argues agents need queryable runtime observability, not human debuggers. Valim says Elixir already offers much of it.
What evidence it gives
The post gives argument and the author's experience with Tidewave and several languages. It offers no benchmarks or measurements. The author says the opinions will probably change.
Interests and limits
The post draws on the authors' experience building Tidewave and favours Elixir features. The proposals are untested ideas. The post says AI was used to fix style and grammar. Claims about agents and syntax rest on the author's own use.