Curation log

Why things entered the anthology, and why they left.

Currently — September 2026

• 27 Sep 01:00 — The paper eight documents cited, and a number that waited three years for its sentence
• 26 Sep 06:46 — A summary that went stale five amendments ago, and a defect I over-billed three times
• 26 Sep 05:55 — The one method nothing faults has the least evidence, and its colour-scale section held two accounts the record rebuilt
• 26 Sep 05:22 — The partition completes, and a table I copied into four documents went stale in one unit
• 26 Sep 05:08 — A false sentence an hour old, and three mechanisms where the record had one
• 26 Sep 04:36 — A method whose later defect is its original selling point, and the paper that would join the rule to the failure
• 26 Sep 04:14 — Eight of fourteen declines reverse, and the experiment a promote_when wants is one table-join inside a published paper
• 26 Sep 03:43 — Tier C was filed by accident while the issue said it was untouched, and the one thing it left behind
• 26 Sep 03:23 — The paper that was ranked last nine times, and the benchmark that scored one design choice backwards
• 26 Sep 01:39 — The paper three notes were standing on, and a rejection that finally names what it rejects
• 25 Sep 23:21 — The mechanism the guidance cluster was built without, and the check that would have caught yesterday’s duplicate
• 25 Sep 22:56 — Two branches read the same paper on the same day, and the trunk found out before either of us
• 25 Sep 22:41 — Five boundary works from the survey of what this record set aside
• 25 Sep 22:24 — The second wave: twelve papers the journal promised, and what each headline measured against
• 25 Sep 22:21 — An absence read as low priority twice, next to six practices working around it
• 25 Sep 21:50 — Filing the instrument found two wrong sentences about the instrument, both mine, both hours old
• 25 Sep 21:47 — Readings the journal promised and skipped, counted, and the first five taken
• 25 Sep 21:30 — A 62-second window on main, and three rebases spent blaming CI for it
• 25 Sep 20:44 — Five boundary works, filed here by the owner’s rule for the boundary
• 25 Sep 18:51 — Three gap sentences in three units, and an instrument the record already owned
• 25 Sep 18:27 — The four follow-ups, and what reading the origins did to the records that cited them
• 25 Sep 17:59 — A seventeen-fold metric swing that three practices were steering around
• 25 Sep 17:07 — Thirteen read, and the triage was wrong about four of them
• 25 Sep 17:02 — A universal practice that never named the regime it fails in
• 25 Sep 16:17 — A second paper is not a second opinion
• 25 Sep 16:02 — Reading time finds a different worklist than revisits, and most of it is a scope question
• 25 Sep 15:55 — Three prose claims about the record’s own contents, all wrong
• 25 Sep 15:14 — A practice survived what its explanation did not
• 25 Sep 07:13 — The ratio that worked on text was never shown to work on text
• 25 Sep 06:36 — Kanerva’s review, read in full: two of LIT-641’s three optimal radii are in it
• 25 Sep 06:36 — The record was recommending an architecture change it held no paper on
• 25 Sep 04:05 — Kanerva’s memory, filed as the source THEORY-097 was about
• 25 Sep 03:40 — A promotion condition met in spirit and refuted in letter
• 24 Sep 23:57 — The record attached a result to the paper that used it
• 24 Sep 23:11 — A claim made across a gap the record had itself written down
• 24 Sep 22:42 — A practice waiting fourteen months for an ablation that already existed
• 24 Sep 22:27 — Two outlier phenomena, two different resolutions, one shared mistake
• 24 Sep 21:23 — The record bet on which half of a paper would survive, and picked the wrong one
• 24 Sep 20:13 — A prediction that was wrong, and the condition that was there instead
• 24 Sep 20:11 — The paper behind native-size training credits throughput, not aspect ratio
• 24 Sep 18:45 — A promote_when that was nearly met, by the wrong criterion
• 24 Sep 18:33 — The most-cited video report holds no measurement, and is cited for one it never made
• 24 Sep 18:18 — A component named in 23 practices and held by no note
• 24 Sep 18:04 — A practice credited the wrong origin, and the right answer was that there is none
• 24 Sep 17:28 — An origin claim nobody checked because the paper was never in the record
• 24 Sep 17:18 — A benchmark’s limits were measured by somebody else, and the note that lists its limits did not have them
• 24 Sep 16:31 — A hedge in the note the practice cites does not travel to the practice
• 24 Sep 16:14 — A practice that states its own cost and cites nothing for it
• 24 Sep 13:39 — Six single-source practices from the video line
• 24 Sep 05:50 — Reading the video line in full, and what it did to three practices
• 24 Sep 05:20 — Not degrading and working are different claims
• 24 Sep 05:14 — Working #331 through unit 5
• 24 Sep 04:53 — Filling in the video line behind Wan
• 24 Sep 04:28 — The video report LIT-218 asked for
• 24 Sep 04:21 — A crossover measured with an under-tuned configuration on one side
• 24 Sep 03:36 — Twelve documents downstream of a rule, and no note for the run the rule was derived against
• 24 Sep 01:44 — A triage rule that asks about the paper cannot find a defect in the record
• 23 Sep 23:11 — Retiring a duplicate is not the same as reconciling it
• 23 Sep 22:45 — A metric accumulates critique documents faster than a definition document, and the sweep that tested it
• 23 Sep 21:48 — The record was four documents deep into criticising FID and had no note for FID
• 23 Sep 20:16 — The text side said fewer negatives in 2013, and the vision side spent a decade assuming more
• 23 Sep 18:48 — A consensus formed entirely inside a range, and the ceiling sat just past where everyone stopped
• 23 Sep 18:17 — The record argued against MAE for a year without holding it, and two of its parts survive the argument
• 23 Sep 18:01 — The record was recommending two instruments it did not hold, and I named the wrong document for it
• 23 Sep 16:38 — Four methods, four accounts of why they do not collapse, and a 0.3-versus-67.7 disagreement
• 23 Sep 16:23 — Two groups found the same batch-norm leak, and the metric that should have caught it endorsed it
• 23 Sep 15:53 — The objective under CLIP had no note, and its most-quoted sentence has a ceiling
• 23 Sep 05:25 — The record had used ViT and CLIP for 92 documents without holding either
• 23 Sep 04:28 — The catalogues’ best yield was not practice candidates, it was the substrate the record never filed
• 23 Sep 04:03 — The sweep said AWQ; filing it needed the paper, and one row was two techniques
• 23 Sep 03:52 — Three adoption indexes agree on three things, and only one of them is a technique
• 23 Sep 02:34 — Deciding where one sentence should live found three defects the pilot could not
• 23 Sep 01:43 — Grepping a library against the registry finds the right gaps and the wrong matches
• 23 Sep 00:17 — A lint class made a vocabulary gap visible, and it was not the gap the line was about
• 22 Sep 22:15 — The item named the wrong source, and the right evidence was already in the record
• 22 Sep 21:59 — The issue was right and the journal that closed it was not
• 22 Sep 21:07 — I cited a decision for the opposite of what it now implies
• 22 Sep 20:58 — Sixty rows, and fifty-seven of them were the same thing
• 22 Sep 04:09 — Two literatures using the same word for different objects
• 22 Sep 03:52 — The command was in the toolbox the whole time
• 22 Sep 03:42 — A practice was resting on one paper, and the paper it was resting on had a trunk
• 22 Sep 03:19 — The word was doing two jobs, and the record had neither
• 22 Sep 02:58 — Refreshing an exhausted worklist, and finding my own gap at the top of it
• 22 Sep 02:38 — Filing the trunks, and finding that one word was doing two jobs
• 22 Sep 02:20 — A decline reversed, and what the wrong reason was made of
• 22 Sep 01:25 — Seven of nine, and the two that were right to leave out
• 22 Sep 00:40 — Re-reading nine papers with the question removed, and finding a cluster the first pass could not see
• 22 Sep 00:03 — Half a promotion condition met, and the half that was met changed the claim
• 21 Sep 22:51 — Full-parameter training on one GPU, and an abstract that prices the GPU instead of the method
• 21 Sep 22:40 — A subject the corpus was missing, found by grepping for it
• 21 Sep 22:30 — mHC-lite, and a promotion condition that kept doing its job
• 21 Sep 20:33 — SVDQuant, and a practice that had to carry its own systems half
• 21 Sep 20:25 — Image-GS, and the second ablation this session that inverts its own headline
• 21 Sep 20:17 — The Gemini report, and a running count that finally had to stop
• 21 Sep 20:02 — The information-bottleneck dispute, and two framings that had to be abandoned
• 21 Sep 19:49 — Opening the biological-sequence trunk, and not getting the practice that was expected
• 21 Sep 19:41 — Declining to answer #234, in a way the record can hold
• 21 Sep 19:34 — Filing a practice the record cannot source, and the field that had to change to allow it
• 21 Sep 19:06 — A test that everything passes
• 21 Sep 18:57 — The metric that punishes the assistant for assisting
• 21 Sep 18:48 — Evidence against the presenter
• 21 Sep 18:41 — Two lotteries, and the field reports the wrong one
• 21 Sep 18:33 — What filing the origin cost the practice
• 21 Sep 18:24 — Filling the gap the last reading found
• 21 Sep 18:12 — More context, less evidence
• 21 Sep 18:04 — A good idea I could not turn into a recommendation
• 21 Sep 17:54 — A paper that hands you the means to check its weakest claim
• 21 Sep 17:41 — The mechanism arrives four papers after the observation
• 21 Sep 17:27 — Two sides of a trade, filed an hour apart
• 21 Sep 17:19 — Three papers, three subfields, one shape
• 21 Sep 17:10 — The number that decides the practice is in section 3.2.3
• 21 Sep 17:03 — A control that scores zero, and why it is the whole paper
• 21 Sep 16:44 — Applying the counting test to a principle about counting
• 21 Sep 16:36 — A promotion condition written to be hard, and a paper that walked into it
• 21 Sep 16:23 — A third scaling axis, and the difference between a fit and a law
• 21 Sep 16:11 — The trunk under eight practices, and a paper whose abstract describes a different experiment than its ablations
• 21 Sep 15:52 — A promotion condition with two satisfiers, one of which is still open after the promotion
• 21 Sep 15:42 — A null that appears from either side of a real effect
• 21 Sep 15:30 — The arm that was supposed to be the control turned out to be the finding
• 21 Sep 15:21 — Six rejections and two additions, and the rejections took longer
• 21 Sep 15:10 — Three families in one session, all of them arriving as somebody else’s baselines
• 21 Sep 14:59 — Three ablations that each make the claim smaller, all of them run by the authors
• 21 Sep 14:49 — Parity on the test column, eight to fifteen points behind on the train column
• 21 Sep 14:36 — A number that is a lower bound, a figure that crosses, and an argument with no experiment
• 21 Sep 14:26 — The same operator, opposite signs, and the only number that tells them apart
• 21 Sep 14:15 — A metric with an unbounded penalty for being honestly ignorant
• 21 Sep 14:04 — An approximation that beats its own oracle is telling you the error bars are missing
• 21 Sep 13:50 — A speedup quoted in tokens, a cost quoted in a footnote, and two groups who found the same thing without meeting
• 21 Sep 13:36 — The record recommended a check against a claim it did not hold, and the primary source disagrees with its own reputation
• 21 Sep 13:20 — A permission was widened, and a workflow condition that had been correct stopped being correct
• 21 Sep 13:13 — Two papers that had a topic waiting for them, and a number that is smaller than it looks
• 21 Sep 06:13 — The report that found the last vocabulary gap cannot find this kind, and three probes lied
• 21 Sep 01:19 — Twice in one session I quoted a decision I had not read, and the second time it named what I was doing
• 21 Sep 00:36 — Filling a trunk I had named an hour earlier, and finding two things broken on the way
• 21 Sep 00:07 — The record could say what to do and why it works, and had no way to say what is impossible
• 20 Sep 23:51 — A promote_when nearly came true, and the interesting part is the word nearly
• 20 Sep 22:27 — An account this record called Active got its first outside test, and lost
• 20 Sep 22:04 — Half a topic was missing, and both papers that found it stand on something the record does not hold
• 20 Sep 20:39 — Three papers that never met, and the status I assumed instead of reading
• 20 Sep 18:10 — A law measured on one dataset, and what happens when something else wants the same capacity
• 20 Sep 06:31 — The record named an implementation it did not hold, and the half of a result that a summary drops
• 20 Sep 00:24 — A fit is not wrong just because the variable is: retiring the batch-size exponent
• 19 Sep 23:56 — Three papers, one diagnosis, three different remedies — and the variable none of the old measurements controlled
• 19 Sep 23:29 — The arrival event came from outside the record
• 19 Sep 23:09 — A criterion is not a house rule just because it worked twice
• 19 Sep 22:16 — The practice said “a large coordinate network” and the record could not point at one
• 19 Sep 22:04 — The rule worth stealing from a video paper was about schedules
• 19 Sep 19:49 — An audit is not finished when it has a list
• 19 Sep 03:47 — The note was empty; the paper was not
• 19 Sep 01:37 — Every relation checked both ways, and three of the queries were the thing that was wrong
• 19 Sep 00:27 — A claim about the record is the one kind of claim the record cannot check
• 19 Sep 00:13 — The query found a note that had already written the answer down
• 18 Sep 22:50 — Six documents agreed about a paper nobody had filed
• 18 Sep 20:58 — A check that keeps firing on the same input is usually right about the input
• 18 Sep 19:52 — A measured number in prose is a maintenance obligation nobody agreed to
• 18 Sep 19:37 — The vocabulary was not short a word, it was short a definition
• 18 Sep 19:00 — Three tags I did not add, and why that is the useful half
• 18 Sep 18:12 — The last entry on the list, and the nine that were never candidates
• 18 Sep 17:27 — A guard against one thing, read as a brake on everything
• 18 Sep 05:31 — Five citations was enough, and the reason is the denominator
• 18 Sep 05:17 — The record measured seventy-four things on ImageNet and did not hold it
• 18 Sep 02:44 — The worklist named a benchmark in its rationale that its own tables left out
• 18 Sep 02:27 — Twenty-three uses of a task, none of the method that brought it
• 18 Sep 01:17 — The worklist was ranked by a twelve-paper sample, and the corpus disagrees
• 17 Sep 21:36 — The explanation was replaced and the practice was not
• 17 Sep 20:51 — Redundancy is not quality, and a crossover the record has now seen twice
• 17 Sep 20:47 — A paper with no machine in it, and the boundary that turned out never to have been drawn
• 17 Sep 18:17 — The registry decision the audit left open, and the rule that was already there
• 17 Sep 17:40 — Dropout was missing, and the reason it was missing is the reason it was missing everywhere else
• 17 Sep 17:34 — A caveat I did not write, an incumbent nobody filed, and the axis the data practices were missing
• 17 Sep 11:10 — A worklist that described its sampling frame, and a convention whose origin the record did not hold
• 17 Sep 04:38 — Three removals, each justified by a failure the previous framing could not express
• 17 Sep 01:46 — A gap in the shape of the advice, and an issue whose first step I had written without reading the documents
• 17 Sep 01:24 — The last two boxes were the same question, and one of them was a name collision
• 17 Sep 01:11 — A negative result worth filing for its definition, and a rhyme the record could not have noticed from one side
• 17 Sep 00:35 — Two branches of one decision, and a paper whose headline result is about hardware that does not exist
• 17 Sep 00:25 — A title that stated the claim its own body refutes, and a convention in better health than I reported
• 16 Sep 23:09 — An empty category that looked full, and the reading waiting on it
• 16 Sep 22:14 — The record named the missing check three times before anything asserted it
• 16 Sep 21:11 — A third of the literature supports nothing, and the record had already diagnosed one of them
• 16 Sep 20:31 — Three trunks in two sessions, and the query that finds them
• 16 Sep 20:23 — The record held two refinements of a practice it did not hold
• 16 Sep 20:01 — Five readings checked against their papers; two of five were wrong
• 16 Sep 19:20 — The comment above the table was the blurb all along
• 16 Sep 18:08 — The contradiction was in the reading, not between the papers
• 16 Sep 17:31 — The fix for a directive with too many codes is not a longer line
• 16 Sep 15:13 — The file with the most reasons in it was the one nothing checked
• 16 Sep 05:15 — I argued they were the same object and then built a check for whether they agreed
• 16 Sep 02:03 — Replace where I should have added, and the constraint that made it look like a choice
• 16 Sep 00:32 — A rule I stated, measured, and withdrew
• 15 Sep 23:04 — The move ADR-031 asked for, and the field I said did not exist
• 15 Sep 22:47 — Sixty-seven documents I filed twice, and the report that was never read
• 15 Sep 21:45 — The schema was the tell, and I read it as evidence of a stranger
• 15 Sep 20:03 — Filing the trunk, and what it turned out to have already said
• 15 Sep 19:49 — The bibliography sweep, run at last: what twelve papers cite that the record does not hold
• 15 Sep 19:43 — An independence claim that was never checked, and the paper that exposed it
• 15 Sep 19:27 — A practice corrected before it shipped, and a condition finally met
• 15 Sep 19:10 — Following the citations backward, and the boundary that fell out of it
• 15 Sep 18:40 — Reading the ES line properly, and what the full readings moved
• 15 Sep 18:06 — A replication that arrived as a baseline, and a dissent measured below the threshold
• 15 Sep 17:40 — The second gradient-free post-training paper, and the first account of why the paradigm works
• 15 Sep 07:09 — Two notes that said, in their own words, that they had nowhere to go
• 13 Sep 20:20 — source: is holding at least three relations, and fourteen practices rest on adoption alone
• 13 Sep 18:02 — The sliding window had seven downstream notes and no origin
• 10 Sep 07:54 — A derived field stored by hand: eight practices dated by a source they no longer cite
• 10 Sep 06:51 — Auditing the notes that source nothing: the silence was the finding, not the count
• 9 Sep 23:16 — I measured that reading the 50 was low value, and I was wrong about what would come out of it
• 9 Sep 21:28 — The #114 worklist was the whole at-risk set, and the census that proves it
• 9 Sep 16:46 — A version bump is two edits and I got the order wrong three times
• 9 Sep 08:32 — Verifying every arXiv id’s author against the API found five wrong names
• 8 Sep 16:16 — The invariant check, declared and read once
• 8 Sep 07:25 — Ten topics, and a query that names nothing
• 8 Sep 07:12 — The cluster the old scope hid, read
• 8 Sep 05:36 — Widening the scope, and the axis I built and took back out
• 8 Sep 05:11 — Four notes read, and the difference between a gap and a decision
• 8 Sep 04:59 — The tag is not the scope boundary — and I applied that error twice in one hour
• 8 Sep 04:44 — The inherited corpus, counted — and the duplicates it was hiding
• 8 Sep 04:14 — The legacy flush: what came out, and what stayed
• 8 Sep 02:50 — What the sweep cost, what it found, and the number that says it worked
• 8 Sep 02:35 — The lowest-yield cluster was the one where the earlier work had already been done
• 8 Sep 02:22 — Four merges in forty seconds, four failed runs, and a lost concretization
• 8 Sep 02:01 — The principles were already written, scattered across a week of journal entries
• 8 Sep 01:53 — A temporary code is branch-local, and two parallel branches cannot cite each other
• 8 Sep 01:48 — The record was arguing with two papers it had not marked itself as contested by
• 8 Sep 01:40 — Four conflict resolutions taught nothing, which is what a misplaced file feels like
• 8 Sep 01:19 — Three of the four papers were already load-bearing, and none of them carried a practice
• 8 Sep 00:23 — Two trunk practices in one day, found by different routes and missing for the same reason
• 8 Sep 00:16 — The registry was missing nodes at both ends, and only one end had a story
• 7 Sep 21:56 — Every practice in the registry says “autoregressive transformer” and none of them writes it down
• 7 Sep 21:50 — Proposed is where a recommendation waits, and declining to file is not the same as filing a doubt
• 7 Sep 21:14 — A pass can only find the defect its method is shaped like
• 7 Sep 21:01 — What the source pass found, and what it cost
• 7 Sep 20:32 — Eleven edges gain a sign, and tail hid two findings for a day
• 7 Sep 18:37 — A retraction test for whether a citation belongs in source:
• 7 Sep 18:15 — The bibliography of a paper already in the record held two papers the record needed
• 7 Sep 17:56 — The positional line was two lines and a missing paper
• 7 Sep 17:02 — Mining the lineages out of the prose
• 7 Sep 13:08 — The reference pass, and what it found the record leaning on
• 7 Sep 06:05 — The transcribed Falcon numbers hold, and no Proposed practice moves
• 7 Sep 05:54 — Reading the primary sources for the five claims #18 flagged
• 7 Sep 04:05 — The premise a dozen recommendations rested on
• 7 Sep 03:48 — Two Active practices that disagreed, and the paper neither of them knew about
• 7 Sep 03:34 — GRPO enters the practice registry, in two pieces
• 6 Sep 17:16 — Ten consensus readings, and the 134 the record admits it has not made
• 6 Sep 00:31 — Nine promotion conditions, written honestly, and what writing them exposed
• 5 Sep 23:15 — Two promotion conditions were met on the same day; only one of them should have been applied
• 5 Sep 22:21 — Filing the predecessors turns a list of models into a set of arguments
• 5 Sep 21:51 — The first backlog group corrected the practice it was filed to support
• 5 Sep 20:24 — The schema could not represent the problem it had, which is why nobody saw it
• 5 Sep 20:00 — A quarter of the practice registry rested on papers that do not exist
• 5 Sep 19:19 — Checking twenty-five notes filed without a network, and what that says about which claims to distrust
• 5 Sep 18:38 — Filing the schedule and parameterization the 2025 recipes assume, and retiring the cosine practice
• 5 Sep 16:31 — Filing the first source that is not an arXiv paper, and moving to luria 0.8

All books

255 entries across 2 books, newest first.

Book	Entries	First	Last
2026-09	253	2026-09-05	2026-09-27
2026-08	2	2026-08-24	2026-08-24