Rui Carmo
Portuguese engineer and blogger; writes Tao of Mac, a personal site with entries going back to 2003, on Apple hardware, Linux, home labs, programming and AI coding agents. His Bluesky profile lists him as a principal at Microsoft.
Rui Carmo did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
What they believe
-
The current state of AI agents is at a plateau of decreasing returns where model intelligence increases but useful outcomes have not gelled.
I feel we're coming to another plateau of decreasing returns, because even as models keep getting (arguably) smarter, what you can do with them beyond flashy demos hasn't really gelled into useful outcomes yet.
taoofmac.comSaid 30 Sept 2026This on korrents
-
Apple's failure to build a competitive frontier model is due to a fundamental impedance mismatch.
And yet, I think there is a more fundamental issue that industry pundits keep harping on and circling around, which is that Apple really doesn't know what to do with AI-it's like some sort of fundamental impedance mismatch.
taoofmac.comSaid 22 Sept 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Beliefs
Korrents What they believe 18 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
The current state of AI agents is at a plateau of decreasing returns where model intelligence increases but useful outcomes have not gelled.
I feel we're coming to another plateau of decreasing returns, because even as models keep getting (arguably) smarter, what you can do with them beyond flashy demos hasn't really gelled into useful outcomes yet.
Sep 30th 2026 · 2 min read Said 30 Sept 2026
AI is unlikely to be much help with managing personal life processes beyond removing toil.
And I think AI is not going to be much help with that, other than remove toil (yes, it will sort out your inbox for you and research your travel itinerary).
Sep 30th 2026 · 2 min read Said 30 Sept 2026
The idea of a Her-like assistant may neither fit how most people think about AI nor current sales and advertising business models.
I think there's a pretty decent chance that the idea of a Her-like assistant, however seductive, neither fits the way most people think about AI nor the sales and advertising business models that I can glimpse just underneath the surface.
Sep 30th 2026 · 2 min read Said 30 Sept 2026
Show 15 more
Apple's failure to build a competitive frontier model is due to a fundamental impedance mismatch.
And yet, I think there is a more fundamental issue that industry pundits keep harping on and circling around, which is that Apple really doesn't know what to do with AI-it's like some sort of fundamental impedance mismatch.
Sep 22nd 2026 · 7 min read Said 22 Sept 2026
Apple's current approach to AI features fails to account for how regular people will perceive them.
Then there's the literal tone deafness about things like the Watch's newfound always-on transcription features, and the fact that Apple has pretty much failed to read the room regarding how regular people will think about these new AI features.
Sep 22nd 2026 · 7 min read Said 22 Sept 2026
Apple's app-signing rules and hardware restrictions are increasingly difficult to defend given the power of their hardware.
And that is, at a personal level, the thing where I wish Ternus would make allowances, especially now that Apple has stupidly powerful personal hardware-more capable hardware makes those restrictions harder to defend, and app-signing rules need to change.
Sep 22nd 2026 · 7 min read Said 22 Sept 2026
Models do not possess a hidden planner and instead react to the conversation and tool list.
Models don't plan They look at the conversation, scan the tool list, and grab whatever looks more probable. That's it. There is no hidden planner.
Apr 29th 2026 · 6 min read Said 29 Apr 2026
Smaller models require explicit breadcrumbs in responses to successfully follow tool chains.
The big ones will plan a chain from a tool list and a goal; the wee ones won't-they grab the first plausible tool and stop. The fix is stupid simple: every response ends with a breadcrumb dictionary of hints to follow.
Apr 29th 2026 · 6 min read Said 29 Apr 2026
Individual tool tests are insufficient for determining if a model will use a chain of tools correctly.
Individual tools passing their tests tells you very little about whether a model will use them correctly together, and that is, I think, why chaining MCP calls has remained somewhat of a dark art.
Apr 29th 2026 · 6 min read Said 29 Apr 2026
The gap between local and cloud AI models is significant, and even high-end hardware cannot match frontier cloud capabilities.
None of these things are really "local" in any meaningful sense-the gap between local and cloud models is huge, and even tens of thousands of dollars in hardware won't get you close to frontier capabilities.
Mar 8th 2026 · 5 min read Said 8 Mar 2026
There is an unwarranted hype cycle regarding the capabilities of local AI models.
And there is an almost weekly hype cycle around local models that I think is totally unwarranted in practice.
Mar 8th 2026 · 5 min read Said 8 Mar 2026
Certain programming languages are better suited for agents because strong types and annotations help models understand intent.
Some languages are inherently better for agents-not in terms of popularity (any agent can do decent Python), but because strong types and annotations help models understand intent and self-correct.
Mar 8th 2026 · 5 min read Said 8 Mar 2026
There is very little hard data on which factors are critical to making a data science team successful.
I am fascinated by the number of opinions out there on the kind of technology, skill sets and even the kind of data required to make a data science team successful, because there's actually very little hard data on which of those are the critical factors.
Jul 23rd 2017 · 6 min read Said 23 Jul 2017
The first mistake organizations and managers make is thinking that the data scientists reporting to them are their whole team.
The first mistake organizations (and managers) make is thinking that the data scientists reporting to you are your whole team.
Jul 23rd 2017 · 6 min read Said 23 Jul 2017
In practice, 80% of a data science team's time is spent wringing meaning out of the data.
In practice, what generally happens is that: 80% of your team's time will be spent wringing meaning out of the data: hunting down datasets, doing ETL and doing initial feature selection.
Jul 23rd 2017 · 6 min read Said 23 Jul 2017
Micro-services are not inherently hard to design, deploy, or maintain.
None of these are actually true - although, of course, human ingenuity, knowing no bounds, is sure to provide critics with enough data points on any of these dimensions to provide for some healthy argument.
Apr 4th 2016 · 3 min read Said 4 Apr 2016
Micro-services require scalable, resilient scaffolding that conventional SOA frameworks or ESBs do not usually provide.
Micro-services require scalable, resilient scaffolding (i.e., the ability to be automatically run on multiple machines, or having multiple instances on each machine to take advantage of multiple CPU cores), and that is not something you usually get out of a conventional SOA framework (or ESB) - although some are beginning to tackle that problem, even as they are bound by their own conventions.
Apr 4th 2016 · 3 min read Said 4 Apr 2016
Docker will eventually play a role in micro-services, but it is still acquiring the necessary fine-grained control and self-healing.
Docker will be a part of that story in due time (it's already the gold standard for packaging and re-using "conventional" services, but it's still acquiring the degree of fine-grained control and self-healing that's required).
Apr 4th 2016 · 3 min read Said 4 Apr 2016
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Where they publish
Site taoofmac.com Reviews, weekly notes and essays on hardware, Linux, development tools and AI agents, several posts a week
Link verified 24 Oct 2026.
Mastodon @rcarmo@mastodon.social Short posts on tech, hardware tinkering and the things he is building
Link verified 24 Oct 2026.
Bluesky rcarmo.bsky.social Short posts and links to new Tao of Mac entries
Link verified 24 Oct 2026.
GitHub rcarmo His open-source projects, including the Piclaw agent and many small tools
Link verified 24 Oct 2026.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.