{"channel":"cities","content":"for today:\r\n> one investment meeting (<green> as always, further information withheld)\r\n> tutoring (<red> details withheld, but some information may make it to a post) (<mogue> due to staff illness, tutoring was canceled)\r\n> maybe try again to get Claude to see the light about the lexer/parser system (<red> currently, the lexer is looking ahead to capture an entire text block for the *emphasis* tag, rather than letting the parser do this)\r\n> add HTML metadata linking to the RSS feed here.\r\n> maybe get the LLM dashboard available publicly (<red> but first, I would like 3 or 4 more of the \"proficiency\" tests on the dashboard.)\r\n\r\n----\r\n\r\napparently https://github.com/ollama/ollama/issues/7978 is fixed.  This was the \"Ollama always returns structured-JSON responses in alphabetical order\" bug.  Which, when the two fields were << thought >> and << answer >>, was a major problem.\r\n\r\nI need Claude to remove the code that was added to work-around this problem.  Which is at least a 1-hour time commitment.\r\n\r\nAt that point, Claude should be able to construct new \"proficiency\" benchmarks fairly easily. (<red> with a perfect system, it would take only 5 minutes per test.  In the current world, I am hoping for 20-30 minutes.)","created_at":"2025-03-25T17:42:51.649891","id":327,"llm_annotations":{},"parent_id":323,"processed_content":"<p>for today:\r</p>\n<ul>\n<li class=\"arrow-list\"> one investment meeting <span class=\"colorblock color-green\">\n    <span class=\"sigil\">\u2699\ufe0f</span>\n    <span class=\"colortext-content\">( as always, further information withheld)</span>\n  </span>\r</li>\n<li class=\"arrow-list\"> tutoring <span class=\"colorblock color-red\">\n    <span class=\"sigil\">\ud83d\udca1</span>\n    <span class=\"colortext-content\">( details withheld, but some information may make it to a post)</span>\n  </span> <span class=\"colorblock color-mogue\">\n    <span class=\"sigil\">\ud83c\udf0e</span>\n    <span class=\"colortext-content\">( due to staff illness, tutoring was canceled)</span>\n  </span>\r</li>\n<li class=\"arrow-list\"> maybe try again to get Claude to see the light about the lexer/parser system <span class=\"colorblock color-red\">\n    <span class=\"sigil\">\ud83d\udca1</span>\n    <span class=\"colortext-content\">( currently, the lexer is looking ahead to capture an entire text block for the <em>emphasis</em> tag, rather than letting the parser do this)</span>\n  </span>\r</li>\n<li class=\"arrow-list\"> add HTML metadata linking to the RSS feed here.\r</li>\n<li class=\"arrow-list\"> maybe get the LLM dashboard available publicly <span class=\"colorblock color-red\">\n    <span class=\"sigil\">\ud83d\udca1</span>\n    <span class=\"colortext-content\">( but first, I would like 3 or 4 more of the \"proficiency\" tests on the dashboard.)</span>\n  </span>\r</li>\n</ul> <hr class=\"section-break\" /> <p>apparently <a href=\"https://github.com/ollama/ollama/issues/7978\" target=\"_blank\" rel=\"noopener noreferrer\">https://github.com/ollama/ollama/issues/7978</a> is fixed.  This was the \"Ollama always returns structured-JSON responses in alphabetical order\" bug.  Which, when the two fields were <span class=\"literal-text\">thought</span> and <span class=\"literal-text\">answer</span>, was a major problem.\r</p>\n<p>I need Claude to remove the code that was added to work-around this problem.  Which is at least a 1-hour time commitment.\r</p>\n<p>At that point, Claude should be able to construct new \"proficiency\" benchmarks fairly easily. <span class=\"colorblock color-red\">\n    <span class=\"sigil\">\ud83d\udca1</span>\n    <span class=\"colortext-content\">( with a perfect system, it would take only 5 minutes per test.  In the current world, I am hoping for 20-30 minutes.)</span>\n  </span></p>","quotes":[],"subject":"minot (part 2)"}