Depender demasiado de la IA para programar puede hacer que perdamos habilidades que luego necesitamos para revisar su código.
#programacion#codereview#ia#desarrollo
Episodio completo:
I built consensus-review: a GitHub Action where two LLMs review your PR independently, and a judge keeps only the findings they agree on.
Running on @gmi_cloud inference.
#GMICloud#GMIStudio#BuildInPublic
One exception: a blocker stays if the judge confirms it from the diff on its own.
In the live demo it caught a removed null guard, a SQL injection and a missing await.
Inline comments, a live summary, token usage + cost, and a local CLI.
#GitHubActions#CodeReview#AI
One pull request got 50+ rounds of AI review in a single night. An AI flagged it, an agent fixed it, repeat.
Here's how a team where AI writes 90% of the code decides what still needs a human.
Full breakdown:
#AI#CodeReview
AI review is a tool. Engineers still own the code, and trust is still part of the job.
Use AI to review code. Not to replace trust.
Do you have this experience before?
#SoftwareEngineering#AI#CodeReview#BackendEngineering
Revisar el código generado por una IA con un modelo distinto ayuda a detectar errores que el primero pasa por alto.
#programacion#codereview#ia#desarrollo
Episodio completo:
GitHub can now remember your whitespace view across pull requests.
Try "Hide for this diff only" to reduce formatting noise. Then show whitespace before approving. Python indentation can change behavior.
#GitHub#CodeReview
Analizamos el uso de agentes de IA como revisores de código y hasta dónde conviene confiar en ellos.
#programacion#codereview#ia#agentes
Episodio completo:
Hablamos de la idea de software factory y de qué papel tiene la revisión de código cuando gran parte del trabajo lo genera la IA.
#programacion#codereview#ia#softwarefactory
Episodio completo:
Revisar grandes cantidades de código generado por IA cansa, y esa fatiga hace más fácil que se cuelen errores.
#programacion#codereview#ia#desarrollo
Episodio completo:
GitHub Copilot code review adds a "Max" effort level (public preview) for deeper reviews of high-severity issues.
Copilot approvals are now generally available, so teams can merge with more confidence.
#GitHub#Copilot#CodeReview
🔗
Code review for $1.00 in / $2.50 out per 1M tokens on @AntSeed.
It catches SQL injection and leaked keys, and it won't invent bugs in clean code.
Peer: 2e301cf63ae792014e4a0d87a17e35c04188e3f2
#CodeReview#AppSec#DevTools#AI#LLM
Welcome to #CodeReview - our new series where devs discuss the inspiration behind their code! Today we're talking to @tomverhoeff, founder of @Methylium and #Windows10 expert, about his BuildCast app. Follow along in the replies below! @GitHub#MSBuild
500 lines? LGTM in 30 seconds. 5 lines? 47 comments about a variable name.
That's Parkinson's law of triviality: the smaller the change, the louder the bikeshed.
中文:500 行 PR 30 秒过审;5 行 PR 却有 47 条评论吵变量名。改动越小,越容易在细枝末节上吵翻天。 #CodeReview