anton kundenko ยท writingessays and notes on vector databases, simd, allocators, and tiny vms.https://anton.kundenko.dev/endatalog on a columnar enginehttps://anton.kundenko.dev/writing/datalog-on-a-columnar-engine/https://anton.kundenko.dev/writing/datalog-on-a-columnar-engine/a new query surface landed this week. eav triples, semi-naive evaluation, and a fixpoint that fits the morsel pipeline naturally.Wed, 22 Apr 2026 00:00:00 GMTrayforce2datalogquery-languagewriting down what i knowhttps://anton.kundenko.dev/writing/writing-down-what-i-know/https://anton.kundenko.dev/writing/writing-down-what-i-know/the experiment teidedb is built on. eighteen years of intuition into a spec dense enough that something else can deliver the implementation.Sun, 12 Apr 2026 00:00:00 GMTteidedbrayforce2agentsmethodologymorsel-driven, finallyhttps://anton.kundenko.dev/writing/morsel-driven-finally/https://anton.kundenko.dev/writing/morsel-driven-finally/rayforce2 with the teide pipeline underneath rayfall. morsels make cancellation, fusion, and profiling structurally simpler.Thu, 12 Mar 2026 00:00:00 GMTrayforce2performanceexecutionmerging teide with rayforcehttps://anton.kundenko.dev/writing/merging-teide-with-rayforce/https://anton.kundenko.dev/writing/merging-teide-with-rayforce/the parallel research project and the production system are now one repo. keeping the language. swapping the engine.Fri, 20 Feb 2026 00:00:00 GMTrayforcerayforce2teidedbarchitecturea week of ui frameworkshttps://anton.kundenko.dev/writing/a-week-of-ui-frameworks/https://anton.kundenko.dev/writing/a-week-of-ui-frameworks/seven days, four desktop ui frameworks, and the decision to keep the database UI close to the engine.Wed, 28 Jan 2026 00:00:00 GMTrayforceuiretrospectivethe day rayforce openedhttps://anton.kundenko.dev/writing/the-day-rayforce-opened/https://anton.kundenko.dev/writing/the-day-rayforce-opened/five years of thinking, two and a half years of private commits. today rayforce is open source.Wed, 26 Nov 2025 00:00:00 GMTrayforcemilestonelooking back at o, ten years inhttps://anton.kundenko.dev/writing/looking-back-at-o-ten-years-in/https://anton.kundenko.dev/writing/looking-back-at-o-ten-years-in/ten years of theplatform. what i think now that i did not think then.Sat, 15 Nov 2025 00:00:00 GMTtheplatformoretrospectiveshelving axlhttps://anton.kundenko.dev/writing/shelving-axl/https://anton.kundenko.dev/writing/shelving-axl/four months, one hundred and seventy-nine commits, then closed. what the experiment gave back and what it did not.Fri, 31 Oct 2025 00:00:00 GMTrayforceaxldesignwhat simdbench taught mehttps://anton.kundenko.dev/writing/what-simdbench-taught-me/https://anton.kundenko.dev/writing/what-simdbench-taught-me/a single c file, a makefile, and a csv. the rule that emerged after four months of measuring.Tue, 23 Sep 2025 00:00:00 GMTrayforceperformancesimdbenchmarkingaxl is rayforce minus the databasehttps://anton.kundenko.dev/writing/axl-is-rayforce-minus-the-database/https://anton.kundenko.dev/writing/axl-is-rayforce-minus-the-database/axldb is the rayforce kernel with the database parts removed to test where the engine boundary actually sits.Sat, 07 Jun 2025 00:00:00 GMTrayforceaxlcdesignbindings before usershttps://anton.kundenko.dev/writing/bindings-before-users/https://anton.kundenko.dev/writing/bindings-before-users/the repo is still private. i am writing the python orm anyway. it is the cheapest way to find out which parts of the c api are wrong.Fri, 18 Apr 2025 00:00:00 GMTrayforcepythonapidesignper-thread heaps, no mutex on the hot pathhttps://anton.kundenko.dev/writing/per-thread-heaps/https://anton.kundenko.dev/writing/per-thread-heaps/every task carries its own allocator. cross-thread frees never touch a lock.Wed, 15 Jan 2025 00:00:00 GMTrayforcecperformanceallocatorconcurrencythe buddy allocator that finally workshttps://anton.kundenko.dev/writing/the-buddy-allocator-that-finally-works/https://anton.kundenko.dev/writing/the-buddy-allocator-that-finally-works/a u64 bitmap, a mask, a count-trailing-zeroes. the smallest fitting order in two instructions.Wed, 04 Dec 2024 00:00:00 GMTrayforcecperformanceallocatorksimple is a tutorial in compressionhttps://anton.kundenko.dev/writing/ksimple-is-a-tutorial-in-compression/https://anton.kundenko.dev/writing/ksimple-is-a-tutorial-in-compression/a week reading kparc's annotated minimal k. two ideas i stole and one i did not.Thu, 11 Jul 2024 00:00:00 GMTrayforcerayfallkdbsix grammars in under nine hundred lineshttps://anton.kundenko.dev/writing/five-languages-in-six-hundred-lines/https://anton.kundenko.dev/writing/five-languages-in-six-hundred-lines/an audit of every peg grammar shipped under std/. the small line count is evidence for grammars-as-values.Sat, 08 Jun 2024 00:00:00 GMTtheplatformparsingpegdslwhat rayfall takes from k, and what it does nothttps://anton.kundenko.dev/writing/what-rayfall-takes-from-k/https://anton.kundenko.dev/writing/what-rayfall-takes-from-k/a year in. writing down which of k's choices made it into rayfall and which i deliberately left behind.Mon, 29 Jan 2024 00:00:00 GMTrayforcerayfallkdbdesignarity-sharded crateshttps://anton.kundenko.dev/writing/arity-sharded-crates/https://anton.kundenko.dev/writing/arity-sharded-crates/tachyon's workspace is one crate per arity class. the cleanest split i have tried.Tue, 19 Sep 2023 00:00:00 GMTtheplatformarchitecturerustthe first monadhttps://anton.kundenko.dev/writing/the-first-monad/https://anton.kundenko.dev/writing/the-first-monad/the kernel signature on day one is the kernel signature today, even though the discipline around it changed.Sat, 19 Aug 2023 00:00:00 GMTrayforcerayfallcdesigninitial commithttps://anton.kundenko.dev/writing/initial-commit/https://anton.kundenko.dev/writing/initial-commit/the rayforce repo went on disk today. seven files, a hundred and three lines, no allocator, no test.Wed, 22 Feb 2023 00:00:00 GMTrayforcectwo atomics, one cursorhttps://anton.kundenko.dev/writing/two-atomics-one-cursor/https://anton.kundenko.dev/writing/two-atomics-one-cursor/the head-tail rewrite. one flag bit makes the difference between "spin once and recover" and "lock to be safe". 0.2.5 ships tonight.Tue, 30 Aug 2022 00:00:00 GMThyperbridgerustconcurrencyatomicsmove, borrow, mixedhttps://anton.kundenko.dev/writing/move-borrow-mixed/https://anton.kundenko.dev/writing/move-borrow-mixed/three laziness modes for join reactions. the runtime knows them. the surface still does not expose them.Tue, 12 Jul 2022 00:00:00 GMTtheplatformconcurrencydesignstarting in c, againhttps://anton.kundenko.dev/writing/starting-in-c-again/https://anton.kundenko.dev/writing/starting-in-c-again/a year and a half of thinking. the database i want to ship lives in someone else's binary, so it has to be in c.Tue, 12 Apr 2022 00:00:00 GMTrayforcecdesignwhat happens when the last receiver dropshttps://anton.kundenko.dev/writing/what-happens-when-the-last-receiver-drops/https://anton.kundenko.dev/writing/what-happens-when-the-last-receiver-drops/two months of using hyperbridge in theplatform. one bug class i had been ignoring is now closed.Mon, 31 Jan 2022 00:00:00 GMThyperbridgerustasyncwakers queue, brieflyhttps://anton.kundenko.dev/writing/wakers-queue-briefly/https://anton.kundenko.dev/writing/wakers-queue-briefly/two days trying to wake n receivers in order. one commit message that is more honest than i would normally write.Fri, 19 Nov 2021 00:00:00 GMThyperbridgerustasyncpost-mortemstarting hyperbridgehttps://anton.kundenko.dev/writing/starting-hyperbridge/https://anton.kundenko.dev/writing/starting-hyperbridge/i needed an mpmc channel with async support. hyperbridge separated that concurrency problem from theplatform.Wed, 17 Nov 2021 00:00:00 GMThyperbridgerustconcurrencyan http request handler in thirty-two lineshttps://anton.kundenko.dev/writing/an-http-server-in-fifty-lines/https://anton.kundenko.dev/writing/an-http-server-in-fifty-lines/composing a socket, a parser, and a reaction across subsystem lines, with no boundary to pay for.Wed, 26 May 2021 00:00:00 GMTtheplatformoparsingconcurrencyone judgment, three subsystemshttps://anton.kundenko.dev/writing/one-judgment-three-subsystems/https://anton.kundenko.dev/writing/one-judgment-three-subsystems/the bet i made in 2016 that i am, finally, fairly sure was right.Sun, 04 Oct 2020 00:00:00 GMTtheplatformoarchitectureafter theplatformhttps://anton.kundenko.dev/writing/after-theplatform/https://anton.kundenko.dev/writing/after-theplatform/four years in production. an embeddable, smaller, c-shaped sibling has been quietly forming.Tue, 15 Sep 2020 00:00:00 GMTrayforcetheplatformdesignjoin patterns instead of mailboxeshttps://anton.kundenko.dev/writing/join-patterns-instead-of-mailboxes/https://anton.kundenko.dev/writing/join-patterns-instead-of-mailboxes/the first reactor is in. join patterns provide atomic matching that actor mailboxes do not.Tue, 17 Mar 2020 00:00:00 GMTtheplatformconcurrencyjoin-calculuschannels are not enoughhttps://anton.kundenko.dev/writing/channels-are-not-enough/https://anton.kundenko.dev/writing/channels-are-not-enough/channel selection works until a transaction needs values from two channels. the reaction has to be the unit.Fri, 30 Aug 2019 00:00:00 GMTtheplatformconcurrencydesigngrammars are values, not stringshttps://anton.kundenko.dev/writing/grammars-are-values-not-strings/https://anton.kundenko.dev/writing/grammars-are-values-not-strings/when grammars become first-class, the language stops being two languages.Wed, 23 Jan 2019 00:00:00 GMTtheplatformparsingpegdesignthe day i gave up on a fix libraryhttps://anton.kundenko.dev/writing/the-day-i-gave-up-on-a-fix-library/https://anton.kundenko.dev/writing/the-day-i-gave-up-on-a-fix-library/a notebook entry from work i cannot share publicly. why grammars-as-values landed for me on this protocol first.Wed, 11 Apr 2018 00:00:00 GMTtheplatformparsingfixno allocation in the inner loophttps://anton.kundenko.dev/writing/no-allocation-in-the-inner-loop/https://anton.kundenko.dev/writing/no-allocation-in-the-inner-loop/the only discipline that has stayed put after a year of writing the o kernel.Tue, 05 Sep 2017 00:00:00 GMTtheplatformperformancerustvec<f64> over a row of structshttps://anton.kundenko.dev/writing/vec-f64-over-a-row-of-structs/https://anton.kundenko.dev/writing/vec-f64-over-a-row-of-structs/a layout note I keep having to reread. why columnar wins by 10x once two passes are involved.Sun, 19 Feb 2017 00:00:00 GMTtheplatformperformancevectorsfirst sketch of ohttps://anton.kundenko.dev/writing/first-sketch-of-o/https://anton.kundenko.dev/writing/first-sketch-of-o/the language has a name. settling the dispatch rule before the syntax.Tue, 08 Nov 2016 00:00:00 GMTtheplatformodesignwhat k got righthttps://anton.kundenko.dev/writing/what-k-got-right/https://anton.kundenko.dev/writing/what-k-got-right/three months in. why arity-classed vector evaluation still gives modern systems a useful design challenge.Wed, 22 Jun 2016 00:00:00 GMTtheplatformkdbvectorsdesignstarting somethinghttps://anton.kundenko.dev/writing/starting-something/https://anton.kundenko.dev/writing/starting-something/kicking off theplatform - a new system, a small language, and three ideas to fit into one runtime.Mon, 14 Mar 2016 00:00:00 GMTtheplatformdesign