<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> <span class="h-card" translate="no"><a href="https://akko.erincandescent.net/users/erincandescent" class="u-url mention">@<span>erincandescent</span></a></span> i am not a fan of message brokers but i did like the policy that any dropped messages get sent to a human operated address</p>

<p><span class="h-card" translate="no"><a href="https://akko.erincandescent.net/users/erincandescent" class="u-url mention">@<span>erincandescent</span></a></span> <span class="h-card" translate="no"><a href="https://mastodon.social/@tef" class="u-url mention">@<span>tef</span></a></span> oooh, nice idea</p>

@tef@mastodon.social this reminds me of how one of the best things a past employer has ever done is making it trivial for the computers to raise tickets for humans Don't know what to do in this edge case? Explicitly punt it to an engineer for now

<p>because this is mastodon, the edge case is this:</p><p>merging the branches together works when the source repos are non overlapping. it also works if you have two repos with a common history, and those repos do not have any other init commits.</p><p>however, if you have some repo that merges in another repo, and try to weave them together, you can&#39;t</p><p>as you&#39;re effectively asking it to merge two different histories for a single repo, and that just doesn&#39;t work out</p>

<p>i spent a few days thinking &quot;how am i gonna handle this weird edge case&quot; and eventually i realised &quot;throw a better error message&quot; was the right answer</p><p>on the plus side i got to replace the caveat in the readme with a &quot;it only breaks if you put garbage in&quot; note</p>

<p>catirclogs.org is almost 15 years old at this point, which is about 15 times more than the half-life of a typical VC-backed service. here&#39;s to 15 years more of IRC logs (and one would hope in 2040 we have something better than IRC that everyone collectively uses. one would hope)</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@whitequark" class="u-url mention">@<span>whitequark</span></a></span> Thank you!</p>

<p>(it&#39;ll be done one last-minute VM resize so that it doesn&#39;t run with disk 80% full later, that is)</p>

<p><span class="h-card" translate="no"><a href="https://mastodon.social/@shriramk" class="u-url mention">@<span>shriramk</span></a></span> quite - this boils down to GenAI being convincingly wrong in ways only someone, who could already answer the question at hand, would understand. </p><p>This leaves us in the spot where GenAI remains a useful tool for skilled people, but a disaster for the creation of skilled people.</p><p>I see these negative effects in the workplace already. Talking to other experienced engineers, I hear some consensus that junior engineers are dangerously trusting of the results they get.</p>