🧑💻Software Engineer, 🎸guitarist (sort of 😝) and lover of 🚵🏂🪁. Cofounder at @CedalioTech. Previously CTO & Cofounder at @wolox and @SyrmoSkate.
Joined April 2009
- Tweets 8,594
- Following 666
- Followers 794
- Likes 3,270
475 Photos and videos
The funny thing about "agentic engineering" is that some of the enterprise practices that never provided too much value are now more useful and easier to maintain.
E.g: I am using a lot of sequence diagrams, flow diagrams, dependencies diagrams when iterating specs with agents.
1
1
62
Jun 15
Agents like to do a mess in the test suite, mainly by doing things that are not efficient. After a few month of agentic coding our test suite performance heavily degraded. This it the fourth iteration but now we are back to sane execution times (total wall time 23s)
2
82
Jun 15
Obviously there is the tradeoff money vs engineering time. In this case I think it was worth to move fast first and then fix the mess, as long as it did not require heavy app code refactor (which i didn't)
1
10
Jun 15
Now we have more instrumentation, tooling and CI checks that detect test performance regression plus I made some skills for our code review bots to detect inefficient test patterns in tests. Will see, but this should be future proof.
9
Jun 15
Codex already knows that I use a lot of dictation when talking to it and sometimes either my pronunciation or the voice-to-text transcribes nonsense (and most of the time codex interprets it correctly).
But sometimes it tells me to clarify and this time it explicitly told me ...
1
60
Jun 15
One clarification: I didn’t understand “Alisa wants to introduce you.” If that was dictated text and meant something for this task, tell me what you intended.
26
Jun 15
The Elixir ecosystem needs to do something about mix and mix test. It is incredible slow. It does not make sense to wait around 3s to run a single test, specially if the test runs in 60ms.
27
Jun 13
More arguments not to depend on a single model provider and to use open weight models.
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
1
45
Jun 12
ExUnit / @elixirphoenix should have something like this to help you monitor how your test are running. Got tiered of our slow test suite and start optimizing with some agents. It is so easy to derail your test suite to be super slow
30
Jun 12
Even with agents we currently don't have the bandwidth to move out of GitHub (and probably BlackSmith) as well. It is becoming insufferable to rely on GitHub for anything and don't give the crap argument "too many agents pushing code", GitHub is unsuable.
Jun 12
WTF is going on!!! @github @useblacksmith Huge delays when checking out code. Who's fault is this?
1
119
Jun 12
Our CI setup will be more way more simple with a statefull server, single machine that runs all the jobs instead of this nonsense of individual runners per job that makes maintainability and overall CI performance a nightmare.
36
Jun 12
WTF is going on!!! @github @useblacksmith Huge delays when checking out code. Who's fault is this?
183
Jun 12
Lets go back to the promo videos where you show me the product full screen and maybe the engineer / founder narrating on top of it. No need to have fancy shots of a person giving a motivational speech for a feature launch
25
Jun 12
If there is no more progress with LLM I am already happy with using them to improve tests suites, refactor flaky tests and do tedious refactors.
30