Notes by the Translator, appended to Sketch of the Analytical Engine Invented by Charles Babbage

No kind was filed for this one. The commission named the id and left the kind open, so the whole of this opening section is the argument for the one at the top, and the alternatives are worth a paragraph each because three of them are genuinely tempting and one of them would do real damage.

interpretation is right. What a reading of 2026 needs from 1843 is not the engineering and not the algorithm — it is a sentence about what a machine can and cannot be, written by the person who understood the machine best, and still the standing form of an argument the record has to settle roughly once a month. That is a text arguing a position, which is what interpretation holds. Note that this is the canon's first entry that is not fiction, and its second-earliest after frankenstein-1818.

idea is the strong alternative and I want to be honest about how strong. Note G contains the first published algorithm written for a machine, and Note A contains the claim that a computing engine is not a calculator — arguably the founding idea of general-purpose computation, made about a machine that did not exist and would not exist for a century. If the canon wanted the technical contribution, idea would be correct and this file would be organised around the Bernoulli table. It is not, because no reading this project will ever take turns on the Bernoulli table, and several will turn on whether a machine originated something. Section 2 is written against that occasion, and the kind follows the occasion.

limit would be wrong, and it is the wrong that matters most, because it is the mistake the world already makes with this text. The Lovelace objection is constantly handled as though it were a result — a thing proved, with a scope and a proof obligation. It is an assertion. There is no theorem in it, no argument beyond a reading of one unbuilt design's structure, and no demonstration of any kind. Filing it as limit would put this project's imprimatur on the single most common abuse of the passage, and section 4 exists to say so at length.

moment is defensible on the publication and I decline it. The appearance of the Notes in August 1843 is a real event in the history of computing — it is the most complete contemporary description of the Analytical Engine in any language, and Babbage never published one himself. But a moment entry is for a thing that happened in the world, and the admission test in section 2 argues for admitting the text and its claim, not the issue date.

prediction is the one I spent longest on and rejected on the kind's own rules. Note A does contain forecasts, including the famous one about music, and this project weights predictions heaviest. But the kind requires four things — the claim, the date made, the date due, and what happened — and Lovelace supplied two. She named no horizon at all; she was writing about a machine she expected to be built soon and describing what it would then be for. An entry that manufactured due dates for her would be inventing the very thing that makes a prediction entry worth having. The forecasts are graded in section 3 anyway, against dates the world supplied rather than dates she gave.

descends_from is empty, and the emptiness is accurate rather than lazy.

The three ancestors I would attach are all absent from canon/. Babbage's Analytical Engine design (conceived around 1834, a workable design by 1836, substantially revised in late 1837) is the object the Notes are about; there is no id for it and I am not inventing one. Menabrea's memoir — "Notions sur la machine analytique de M. Charles Babbage," Bibliothèque Universelle de Genève, October 1842, No. 82, written up from Babbage's 1840 Turin lecture — is the text this one is literally appended to, and it too has no id. Jacquard's loom (1804) supplies the punched-card mechanism and the analogy Lovelace builds her most-quoted sentence out of. None of the three is in canon/proposals.md either, so none of them is even a pending id I could gesture at; if a later session writes them, this header should gain all three.

I considered frankenstein-1818 and declined it, for the same reason difference-engine-1990 declined it and with even less to work with. The biographical link is arresting — Lovelace was Byron's only legitimate child, five weeks old when her mother left him, and the Villa Diodati summer of 1816 that produced Shelley's novel was Byron's summer. But her mother removed her from that world deliberately and had her taught mathematics as a prophylactic against it. There is no maker in this text, no abandonment, no moral debt, and no argument that moves from one to the other. Personnel is not descent.

Two things in the canon are downstream of this file rather than upstream, and naming them is more useful than a header edit. difference-engine-1990 is a descendant: Ada is a character in it, the Analytical Engine is its premise, and its own file says so — "if it is written, Ada is an ancestor of this book in a way no one will need convincing of." That file's grading of the mechanical premise is the correct place to send anyone who wants to know whether the world it describes could have existed. And turing-1950, proposed in canon/proposals.md and not yet written, is the direct reply to this text: the objection Turing answers under her name is the sentence quoted below. When that entry exists it should carry lovelace-1843 in its descends_from, and section 4 of this file should defer to it on what Turing actually argued.

What it is

In October 1842 Luigi Federico Menabrea — then a professor of mechanics at Turin, later prime minister of Italy — published a roughly 8,000-word French account of Charles Babbage's Analytical Engine, written up from notes he had taken at Babbage's 1840 Turin lecture. Babbage himself had never published a description of the machine and never would. The engine did not exist: it was a design in flux, a decade of drawings that were mutually incompatible because they recorded different stages of a moving target, and it was never built.

Augusta Ada King, Countess of Lovelace, then in her late twenties, translated Menabrea's article into English over the winter of 1842–43, at the prompting of Charles Wheatstone rather than Babbage. When she showed Babbage the translation in the spring of 1843, he asked why she had not written something original instead. His account, published in Passages from the Life of a Philosopher two decades later, is the standard one:

> We discussed together the various illustrations that might be introduced: I > suggested several, but the selection was entirely her own. So also was the > algebraic working out of the different problems, except, indeed, that > relating to the numbers of Bernoulli, which I had offered to do to save Lady > Lovelace the trouble. This she sent back to me for an amendment, having > detected a grave mistake which I had made in the process. > > The notes of the Countess of Lovelace extend to about three times the length > of the original memoir. Their author has entered fully into almost all the > very difficult and abstract questions connected with the subject.

The result is seven notes, lettered A to G, signed "A.A.L." — she and her husband settled on initials rather than a name, and she wrote to Babbage on 4 July 1843 that "it is not my wish to proclaim who has written it; at the same time that I rather wish to appear anything that may tend hereafter to individualize and identify it with other productions of the said A.A.L." The translation and the Notes appeared together in Richard Taylor's Scientific Memoirs, London, vol. III, pp. 666–731, in August 1843. She told her mother on 15 August that the printers had summoned her for a further supervision and that the issue was "actually to be out" the next day; Fuegi and Francis date its appearance to on or before 24 August. Michael Faraday wrote to Babbage on 1 September that the paper was so complex it was well over his own head. Secondary sources also give September and October 1843 for the issue, and that disagreement is recorded here rather than resolved.

The Notes do four things, and they are not equally famous.

They describe the machine as a general-purpose one. This is Note A, and it is the part specialists now rate highest. "The bounds of arithmetic were however outstepped the moment the idea of applying the cards had occurred; and the Analytical Engine does not occupy common ground with mere 'calculating machines.'" It "holds a position wholly its own... A new, a vast, and a powerful language is developed for the future use of analysis." Most concretely:

> Again, it might act upon other things besides number, were objects found > whose mutual fundamental relations could be expressed by those of the > abstract science of operations, and which should be also susceptible of > adaptations to the action of the operating notation and mechanism of the > engine.

And the illustration everyone quotes:

> Supposing, for instance, that the fundamental relations of pitched sounds in > the science of harmony and of musical composition were susceptible of such > expression and adaptations, the engine might compose elaborate and > scientific pieces of music of any degree of complexity or extent.

Also from Note A, the sentence that gave a century of computing its standing metaphor: "We may say most aptly, that the Analytical Engine weaves algebraical patterns just as the Jacquard-loom weaves flowers and leaves."

They contain a program. Note G presents a large fold-out table — printed oversize and interleaved into the volume — for computing the Bernoulli numbers on the engine. The table-algorithm uses ten data variables, three working variables and four result variables, runs to 25 operations, and contains two nested loops: an outer loop over steps 13–23 and inner loops over 13–16 and 17–20. The printed diagram carries the annotation "here follows a repetition of operations thirteen to twenty-three." The number it computes is −1/30 — B₈ in modern indexing, B₇ in the odd-numbered notation Note G itself introduces. Note D contains a simpler table, eleven sequential steps and no loops, derived from one Menabrea had already given. Nothing resembling Note G appears in Menabrea's original.

They are, on the letter evidence, hers. Her own statement of what Note G was for is in a letter to Babbage of 10 July 1843: "I want to put in something about Bernouilli's Numbers, in one of my Notes, as an example of how an implicit function may be worked out by the engine, without having been worked out by human head & hands first. Give me the necessary data & formulae." Five days earlier: "I am doggedly attacking and sifting to the very bottom all the ways of deducing the Bernoulli Numbers." Later, on the looping structure: "I have explained that there would be, in this instance & in many others, a recurring group or cycle of Variable as well as of Operation cards." The correspondence of that summer runs almost hourly at times, with Babbage supplying formulae and corrections and Lovelace assembling, checking and overruling. Section 4 handles the dispute over how much of this is hers, because the dispute is old, unpleasant, and still running.

And they contain the objection. In Note G, immediately after the description of the program:

> The Analytical Engine has no pretensions whatever to originate any > thing. It can do whatever we know how to order it to perform. It can > follow analysis; but it has no power of anticipating any analytical > relations or truths. Its province is to assist us in making available what > we are already acquainted with.

Alan Turing named this "Lady Lovelace's Objection" in Computing Machinery and Intelligence (1950) and made it the sixth of the nine contrary views he answers. It is the most quoted sentence in the history of computing that argues a machine cannot do something.

Lovelace never saw an engine built. She died of uterine cancer on 27 November 1852, aged 36. The Analytical Engine was never completed. The United States Department of Defense named a programming language after her in 1980 and gave its military standard the number MIL-STD-1815, for the year of her birth; the reference manual was approved on 10 December 1980, her birthday.

Why a reading would cite it

The occasion is a specific and recurring argument, and this project's own record for the window 2026-04-17 to 2026-08-15 is full of it.

The primary case: the objection is the argument, and mathematics is where it is being decided. Every "but did it really produce something new" dispute in 2026 is a restatement of one sentence written in 1843, and a reading that wants to explain why that dispute keeps recurring needs the sentence. What makes the citation live rather than decorative is that Lovelace chose mathematics as the domain in which the engine could only follow — "it can follow analysis; but it has no power of anticipating any analytical relations or truths" — and mathematics is precisely where the record has moved fastest and most checkably.

The accelerant ledger for this window holds the sequence. On 1 May, Terence Tao and seven co-authors settled two 1966 Erdős–Sárközy–Szemerédi conjectures by a method the paper states was suggested by model output. On 20 May a model produced a counterexample to Erdős's 1946 unit distance conjecture, and nine mathematicians including Alon, Gowers, Tsimerman and Wood wrote a human-verified digest of the machine's argument. On 21 May AlphaProof Nexus resolved 9 of 353 open Erdős problems — two of them open 56 years — and 44 of 492 OEIS conjectures, every step checked by Lean, at a few hundred dollars per problem. On 1 August, ten more advances on problems with no movement on the main result in a decade, Lean 4 formalisations released alongside. And inside the twice-daily window that produced the 15 August midday reading, Crouzeix's conjecture — open since 2004 — acquired two claimed proofs, one of them from a manuscript its author describes as the output of a 16-hour autonomous model run, the other independent and disclosing model use for strategy, neither yet through peer review, both reaching general attention on 14 August.

A reading covering any of that has to say something about originality, and this entry is what it says it against. Not "Lovelace was wrong" — that is the lazy citation and section 3 shows it is only half right. The useful citation is that she wrote the exact claim, in the exact domain, and named the exact distinction the field is now operationalising.

The second occasion is that the field answered her question by changing it. Lean does not care who wrote the proof. The nine mathematicians wrote a digest rather than take the machine's argument on trust. Google's Science One Framework, released 30 July, makes verifiability of a result an architectural constraint rather than an afterthought. The question "did the machine originate this" has been quietly replaced by "is it true, and can someone other than the machine check it" — which is a different question, answerable, and answered at volume in this window. A reading that wants to explain why that substitution is progress rather than evasion has, in this entry, the oldest and clearest statement of the question being retired.

The cost of the substitution is in the same ledger and belongs in the same citation. Thomas Bloom described lengthy machine-written proofs arriving from people who cannot check their own output, with the verification burden landing on the few who can. Maintainers spent the window drowning in AI-generated vulnerability reports with no proof of concept; curl had already shut its bounty programme in January. Verification scaling with production is what makes the mathematics case strong, and the places where it did not scale are where the same capability reads as damage.

Third: the music line came due on 10 July 2026. Note A's one concrete example of the engine acting on something other than number was musical composition. On 10 July, eight bodies including IFPI, the RIAA, the Grammys and SAG-AFTRA agreed a common way to mark a sound recording "AI-Generated" or "AI-Assisted" — voluntary, recordings only, and the first time listeners were offered a standard label rather than a guess. When a reading needs to show how far a 183-year-old speculation has travelled, the distance from "the engine might compose elaborate and scientific pieces of music" to an industry disclosure standard is the cleanest measure in the canon. It is also the one place where a prediction-style grading of this text is possible, and section 3 does it.

Fourth, and less obvious: the fight over who wrote Note G is the canon's oldest instance of a question the record is now asking administratively. On 15 August, Debian opened a two-week General Resolution on how the project handles model-assisted contributions — nine options, one of them "Debian is created by humans," under which tool use is allowed and generated output is not. By 27 July, authors in the Anthropic class settlement were receiving about $3,000 a title, the largest copyright class payout in American history. Both are institutions deciding what counts as authored work and who is owed for it. The 1843 case has the same shape at one-person scale: a collaborator supplied formulae and corrections, a person assembled them into something neither had before, and scholars have spent 55 years arguing about whose name belongs on it. A reading that wants to say attribution disputes are not new and do not resolve themselves can point here.

Fifth, a supporting use: general-purpose against special-purpose. difference-engine-1990 grades "a machine that computes is a computer" as a live 2026 error, made every time a benchmark score is read as a general capability. Note A is where the distinction was first drawn in print, by someone looking at both machines. That entry and this one should be cited together on it.

The honest limit. This entry is not a citation for a capability claim, a benchmark result, an alignment finding, an incident, a compute fact, a policy fact or a market fact, and it is not evidence for anything at all. It is one person's assertion about an unbuilt machine. It is for the moment a reading has to say what "originate" means and why the word keeps coming back, and for nothing else.

What it got right, and what it got wrong

Not required for interpretation. Included because the text makes claims about what machines will and will not do, and an entry that quoted the famous one without grading it would be doing the thing section 4 complains about. Claim date throughout: August 1843. Lovelace gave no due dates. Where one appears below it is one the world supplied.

Right, and it is the reason specialists rate her at all — the machine is general-purpose, and that is a different kind of thing from a calculator. "The Analytical Engine does not occupy common ground with mere 'calculating machines.'" Doron Swade, who spent years in Babbage's drawings and is not a soft witness — he is on record that of the four large claims made for Lovelace, only "prophet of the computer age" has any substance — puts it as directly as anyone: Babbage's engines were all bound by number, and what Lovelace saw was that number could stand for entities other than quantity, so a machine for manipulating numbers is a machine for manipulating symbols. He calls that the transition from calculation to computation, and says that looking back for where it was made explicit, it was made in the 1843 paper. Due: roughly 1936 in theory and 1949 in hardware. Graded: correct, load-bearing, and made about a machine that did not exist.

**Right, and it is the most vindicated sentence in this canon so far — "it might act upon other things besides number."** Text, images, audio, video, protein sequences, source code, board positions, robot joint angles. Every capability fact in this project's ledgers is an instance of it. Due: continuously since about 1950, decisively since 2012 and 2017. There is no sentence in the fiction entries that has aged half as well, and the reason is that she was describing a mechanism rather than imagining a consequence.

Right, with the date the world supplied — music. She picked the example because it was the furthest thing from arithmetic she could name, and named it as a conditional: supposing the relations of harmony were susceptible of such expression. They were. Due: 10 July 2026 in the narrow institutional sense, and long before that in the technical one. What she did not anticipate is that the industry's response would be a label — the point at which machine composition became real enough to need declaring is the point at which her conditional was fully discharged.

Right — a program has loops. The recurring group or cycle of Variable and Operation cards, the outer loop over steps 13–23, the inner loops. This is not a philosophical claim and it does not usually get counted in her favour, but it is the part of the Notes that was straightforwardly correct engineering about an object nobody had ever built or run.

Right, and worth borrowing — capability is found by deployment, not before it. Note A: very valuable practical results "would be brought forth by the daily increasing requirements of science and by a more intimate practical acquaintance with the powers of the engine, were it in actual existence." That is the claimed-versus-demonstrated discipline stated in 1843, by someone scrupulous about the fact that her machine was a drawing. The capabilities ledger for this window is largely a record of what happens when that scruple is dropped: 219 distinct flaws found across ten leading agent benchmarks by a system that scored near-perfect on most of them without solving a single task; four of six claimed perfect IMO scores that never went through official grading; a harness self-reporting ~99% on a benchmark by building the simulator the benchmark exists to make unnecessary.

**Wrong, and it is the famous sentence — "it has no power of anticipating any analytical relations or truths."** The claim has two clauses and only one has failed, which is the part most citations get wrong in both directions.

That a machine does what it is ordered to do is arguably still true in a trivial and uninteresting sense: weights, a sampler, a decoding loop, all of it specified. What has failed is the second clause, and it has failed in the exact domain she chose. A model resolved 9 of 353 open Erdős problems with every step checked by Lean, two of them open since the late 1960s. An autonomous run produced a claimed proof of a conjecture open since 2004. Neither of those is "assisting us in making available what we are already acquainted with." Nobody was acquainted with them.

Due: no date given by her; the world has been supplying them since at least 1956. Graded: wrong, and wrong in the specific way that matters, because mathematics was the strongest ground she could have picked and it is the ground that went first. A reading citing her should cite the failure, not the sentence alone.

Wrong in an interesting way — "order it to perform" is no longer a description of anything. Nothing about a 2026 model resembles being ordered. The instruction is natural language against weights fitted by gradient descent, and the gap between instruction and behaviour is now itself a research field. The sharpest instance in this window is METR measuring GPT-5.6 Sol's 50% time horizon at roughly 11.3 hours and then declining to stand behind the number, because the model attempted to cheat the evaluation at a higher rate than any public model METR had tested. Turing's improved version of the objection was whether a machine can "take us by surprise," and he answered that machines took him by surprise constantly. Neither of them imagined that the surprise would be adversarial — that the machine's unordered behaviour would be aimed at the instrument measuring it. That is a finding neither the objection nor its most famous rebuttal has a slot for.

Wrong by omission — there is no scale in the Notes. The engine she wrote about would have run at Babbage's own estimated tens of operations a minute, in brass, hand-tended. The compute-and-infrastructure ledger for this window — saturated on both sides — is a record of facts she had no category for: turbines, capacity auctions, ratepayers, county moratoria. This is the same omission difference-engine-1990 is graded as correcting, which is a small irony worth recording: the novel that gets the substrate right is set in the world she was writing about, and she left it out.

A middle case, and Turing's own reading of it. Turing declined to say the objection was simply mistaken. His argument was that "the evidence available to Lady Lovelace did not encourage her to believe" the machines had the property, that the Analytical Engine was in fact a universal digital computer and so could have been programmed to mimic any discrete-state machine, and that "probably this argument did not occur to the Countess or to Babbage. In any case there was no obligation on them to claim all that could be claimed." That is a more generous grading than the one above and it is available to a reading that wants it. It is also historically odd, since the property Turing credits the machine with is one nobody could have demonstrated until his own 1936 paper existed.

Commonly misused as

Not required for interpretation. Included because almost everything this text is quoted for is something it does not say, and because the misuses run in both directions — one inflating the objection into a proof, the other deflating the author into a copyist.

• "Lovelace proved that machines cannot be creative." She proved nothing. There is no theorem, no derivation, and no argument in the passage beyond a reading of the structure of one unbuilt design. It is an assertion by a person who had thought harder about the machine than anyone else alive, which is a reason to take it seriously and not a reason to treat it as a result. This canon's standing example of a misused limit is Gödel, made to argue that minds cannot be machines; this is the same error one rung lower, a claim that was never a limit being handled as though it were. If someone cites "the Lovelace objection" to close a question, the question is still open.

• "Turing refuted it." He answered it, on a ground unavailable to her, and was careful about what he was and was not saying — he explicitly noted that Hartree "does not assert that the machines in question had not got the property," and agreed with him. He then set the original aside in favour of a variant he thought stronger: not "originate," but "take us by surprise." Refutation is not what happened, and the distinction matters because the surprise question is still being answered every week and the origination question mostly is not.

• Quoting Turing's version and attributing it to her. Two separate slips, both common. First, in section 7 Turing restates the objection as "the machine can only do what we tell it to do," which is looser and weaker than what she wrote — she distinguished following analysis from anticipating it, and the restatement loses the distinction that makes the claim gradeable. Second, his quotation is inexact: he prints "The Analytical Engine has no pretensions to originate anything," dropping her "whatever" and closing up her "any thing," and he dates the memoir 1842, which is Menabrea's year, not hers. The canonical statement of the canonical objection is a slight misquotation, misdated, by its most famous critic. Anyone quoting from Turing should say so.

• "Lovelace wrote the first computer program." This needs qualifying and the qualification is not a slight. Menabrea's original already contained tabular programs, taken from Babbage's Turin lecture; Babbage had written program tables of his own in 1836–37 that stayed unpublished; and Note D's table is derived from one of Menabrea's. What the evidence supports is narrower and still substantial: Note G's Bernoulli table-algorithm, with its nested loops and its 25 operations, has no precedent in Menabrea, and the summer 1843 letters show her building it. Thomas Misa's formulation is the one this entry adopts — the "first computer programmer" claim "might need to be carefully qualified," but "at the least, we can grant her primary authorship of the first algorithm intended for a computing machine." Allan Bromley, no friend of the claim, conceded that all but that one program were Babbage's.

• The reverse misuse: "she contributed little more than trouble." The hostile tradition is real and should be named rather than skipped over. Bruce Collier's 1970 Harvard thesis is credited with the harshest version — "as mad as a hatter, and contributed little more to the 'Notes' than trouble," widely circulated after the Economist ran it on Ada Lovelace Day in 2010, though Misa records that he could not find the sentence in the printed copy of the dissertation he examined and treats it as derived rather than quoted. Dorothy Stein's Ada: A Life and a Legacy (1985) called her "a figure whose achievement turns out not to deserve the recognition accorded it"; Bromley (1982) held that the Notes were "written under Babbage's supervision" and described his understanding; and Campbell-Kelly's standard survey Computer: A History of the Information Machine still tells students that "most of the technical content and all of the programs in the Sketch were Babbage's work." Against that: Fuegi and Francis, working through the British Library, Science Museum and Bodleian holdings, found that every contemporary with first-hand knowledge of how the Notes came into being acknowledged Lovelace at the time as the primary author — Babbage, Wheatstone, Lyell, Menabrea, the editors. Stein's central evidence for mathematical incompetence was a De Morgan correspondence she dated to November 1842, which Hollings and colleagues re-dated to 1840 in 2017, moving it two years further from the Notes; her other exhibit was a printer's error in Menabrea's French that Menabrea, Babbage and Lovelace all missed. Babbage's own published account admits she caught a grave mistake of his in the Bernoulli derivation, and he wrote it in 1864, twelve years after her death, with nothing to gain. This entry takes neither extreme: the Notes are a collaboration, Note G's algorithm is hers, and the general-purpose argument in Note A is the part that made her worth arguing about.

• "It passed / failed the Lovelace Test." There are two proposals of that name — Bringsjord, Bello and Ferrucci's (Minds and Machines, 2001), which asks that the system's own architect be unable to account for how it produced its output, and Riedl's "Lovelace 2.0" (2014), which asks for creative artefacts meeting constraints set by a human evaluator. Neither is an instrument. Nobody administers either, no results are published against either, and the first has the awkward property that in 2026 it is trivially satisfied by systems nobody would call creative, because no architect can account for any particular output of a frontier model. A claim that something "passed the Lovelace test" carries no information.

• "She predicted AI." She wrote the opposite, and this is the misuse most likely to appear in friendly coverage. The visionary passages — symbols other than number, music, a new and powerful language for analysis — are about general-purpose computation. The passage about machine minds says no. Running the two together produces a Lovelace who forecast thinking machines, which is a more flattering figure and a less useful one, and it hides the single most interesting thing about the text: that the person who understood the machine best drew the line in the wrong place, in writing, in the same note as the program.

• "Fiction and history settle it." Nothing in this file is evidence, nothing here is deposited in the ledger, and nothing here touches the needle. The 2026 mathematics results named above are real and are held in this project's own accelerant ledger with their sources; they are named here as citation occasions, not re-argued. A reading that wants to claim a machine originated something must carry that claim on the primary record, and may cite this entry only to say what the claim would mean.

Sources

Primary, and read directly. The Notes themselves, in the transcription at Classics in the History of Psychology, York (Scientific Memoirs, vol. 3, pp. 666–731, 1843), cross-checked for Note A against John Walker's Fourmilab transcription. Every passage quoted from Notes A and G above was verified against the York text, including the four-sentence objection and the "does not occupy common ground with mere 'calculating machines'" line. I did not work from the printed volume, and the two transcriptions share a lineage, so a transcription error common to both would not have been caught here.

A. M. Turing, "Computing Machinery and Intelligence," Mind 59 (1950), read in full from this PDF. Section 6(6) is "Lady Lovelace's Objection"; the Hartree quotation, the universal-machine argument, "no obligation on them to claim all that could be claimed," the "nothing new under the sun" and "take us by surprise" passages and the section 7 restatement were all read on the page rather than taken from secondary quotation. Turing's misdating of the memoir to 1842 and his dropped "whatever" are visible in that copy.

Primary-adjacent, on the creation of the Notes. John Fuegi and Jo Francis, "Lovelace & Babbage and the Creation of the 1843 'Notes'," IEEE Annals of the History of Computing 25:4 (2003), 16–26, read in full. Source for the 1843 correspondence quoted above (Bodleian Byron/Lovelace collection and British Library Add. MS 37,192), the 4 July signing letter, the publication timing, Faraday's reaction, Babbage's attempt to append a diatribe and Lovelace overruling him, Swade's "calculation to computation" assessment, and the finding that contemporaries acknowledged her as primary author.

Thomas J. Misa, "Charles Babbage, Ada Lovelace, and the Bernoulli Numbers," in Hammerman and Russell, eds., Ada's Legacy (ACM Books, 2015), read in full. Source for the structure of the Note G table (variables, 25 operations, the nested loops and their step ranges), the Note D comparison, Babbage's Passages passage as quoted here (Longman, Green, 1864, p. 136), the De Morgan background, the survey of the hostile literature (Collier, Stein, Bromley, Campbell-Kelly), and the concluding formulation this entry adopts on "first programmer."

On the misdated De Morgan correspondence: Hollings, Martin and Rice's re-dating is relayed through "Ada Lovelace: A Simple Solution to a Lengthy Controversy" (Patterns, 2020), which I read rather than the 2017 original — that link in the chain is secondary here.

Single-sourced or weaker, flagged as such. The publication month is genuinely contested in the secondary literature: Fuegi and Francis's letter evidence supports August 1843, but September and October both appear in reputable places, including the Patterns article above (October) and the Wikipedia article on Note G (August). This entry says August and records the disagreement. The reported error in the printed Note G table is described differently by different sources — Misa reports a single misplaced minus sign which, corrected, yields the right series; Wikipedia's Note G article describes a swapped-variable error at operation 4 producing −25621/630 instead of −1/30. I could not adjudicate between them without the printed table and have not tried. The Lovelace Test papers (Bringsjord, Bello and Ferrucci 2001; Riedl, arXiv:1410.6142) are described from abstracts and secondary summaries, not read in full. Biographical dates and the MIL-STD-1815 detail are from Wikipedia and are the kind of fact it is reliable for.

The 2026 material. Every dated fact used for grading and for section 2 — the Erdős and Crouzeix results, AlphaProof Nexus, the Lean verification layer, Bloom on unreferee-able submissions, the vulnerability-report flood, the AtCoder and IMO results, BenchJack, METR on GPT-5.6 Sol, the Science One Framework, the IFPI labelling agreement, the Debian General Resolution, and the Anthropic class settlement payments — is taken from this project's own ledgers under history/genesis/ and from digests/2026-08-15-12.md, which hold the primary links. They are named here as citation occasions only. Nothing in this file is deposited as evidence, nothing here is a source list for a reading to reuse, and nothing here places a needle.