<p><span class="h-card" translate="no"><a href="https://fosstodon.org/@kevinbowen" class="u-url mention">@<span>kevinbowen</span></a></span> I realize I don&#39;t really remember those particulars :-) Been about a year ago, I think.</p>

<p>We&#39;ve anchored just off from Jamestown.</p>

<p><span class="h-card" translate="no"><a href="https://chaos.social/@nanographs" class="u-url mention">@<span>nanographs</span></a></span> happy to see the general structure of the design I proposed is still serving you reasonably well</p>

<p><span class="h-card"><a href="https://mastodon.social/@shriramk" class="u-url mention">@<span>shriramk</span></a></span> my understanding of the gossip is “probably lied about training costs because they couldn’t admit to using embargoed NVIDIA tech” and “probably plagiarized heavily from the OpenAI plagiarism machines” but their end result supports <span class="h-card"><a href="https://types.pl/@wilbowma" class="u-url mention">@<span>wilbowma</span></a></span>’s analysis imo even if you just accept all that as true</p>

<p><span class="h-card" translate="no"><a href="https://types.pl/@wilbowma" class="u-url mention">@<span>wilbowma</span></a></span> Yeah, it&#39;s not at all clear what the truth about DeepSeek is.</p>

<p>“Gratitude is due to the heroic Ukrainian soldiers who shed their blood in defense of the values of the free world,” Mr. Walesa, who served as Poland’s first elected president after the collapse of Communism, and other signatories said, adding, “We do not understand how the leader of a country that is a symbol of the free world cannot see this.”</p><p><a href="https://www.nytimes.com/2025/03/03/us/politics/lech-walesa-trump-letter.html?smid=url-share" target="_blank" rel="nofollow noopener noreferrer" translate="no"><span class="invisible">https://www.</span><span class="ellipsis">nytimes.com/2025/03/03/us/poli</span><span class="invisible">tics/lech-walesa-trump-letter.html?smid=url-share</span></a></p>

<p><span class="h-card" translate="no"><a href="https://types.pl/@jeremysiek" class="u-url mention">@<span>jeremysiek</span></a></span> <span class="h-card" translate="no"><a href="https://types.pl/@wilbowma" class="u-url mention">@<span>wilbowma</span></a></span> <span class="h-card" translate="no"><a href="https://types.pl/@sliminality" class="u-url mention">@<span>sliminality</span></a></span> I guess I would prefer not to characterize either injection or projection from Dyn as an up or downcast, but it&#39;s definitely right that TR provides implicit casts to Dyn but not from Dyn.</p>

<p>So we do have a blank command at the end of an aborted frame already, nice. Have some more places to go through to make sure we have blanks everywhere we want to stop milling when done doing something.</p>

<p>Ok so it looks like the only time we are actually blanking is at the end of a pattern. At the end of acquiring an image we are not currently blanking. At the end of an image, the beam is returned to (0,0), the top left corner. You can see that it is digging deep holes there. So whenever I had previously acquired an image before starting a pattern, it ends up digging a hole there.</p><p>The holes are limited in size due to sputter re-deposition ratios that change as the hole gets deeper.</p>