I shipped PatchPitch 0.3.0. Here is what the failed 4B run taught me.

Description and use

<p>The faster run took three minutes. It produced zero drafts I could accept.</p> <p>The slower run took almost twenty-seven minutes. It produced all four, then repeated the completed job in less than a second without asking the model anything.</p> <p>That difference became a useful release test for <strong>PatchPitch 0.3.0</strong>, my open-source tool for turning development facts into announcement drafts for Telegram, X, DEV and GitHub. I shipped it on 24 October. Getting there meant looking beyond whether the model returned some text.</p> <p>In <a href="https://dev.to/drevendev/the-code-wa

Problem

<p>The faster run took three minutes. It produced zero drafts I could accept.</p> <p>The slower run took almost twenty-seven minutes. It produced all four, then repeated the completed job in less than a second without asking the model anything.</p> <p>That difference became a useful release test for <strong>PatchPitch 0.3.0</strong>, my open-source tool for turning development facts into announcement drafts for Telegram, X, DEV and GitHub. I shipped it on 24 October. Getting there meant looking beyond whether the model returned some text.</p> <p>In <a href="https://dev.to/drevendev/the-code-wa

Outcome

<p>The faster run took three minutes. It produced zero drafts I could accept.</p> <p>The slower run took almost twenty-seven minutes. It produced all four, then repeated the completed job in less than a second without asking the model anything.</p> <p>That difference became a useful release test for <strong>PatchPitch 0.3.0</strong>, my open-source tool for turning development facts into announcement drafts for Telegram, X, DEV and GitHub. I shipped it on 24 October. Getting there meant looking beyond whether the model returned some text.</p> <p>In <a href="https://dev.to/drevendev/the-code-wa

Evidence

I shipped PatchPitch 0.3.0. Here is what the failed 4B run taught me.

Source