britt (@p3ery) header
britt (@p3ery) profile picture

britt

@p3ery

I love music, my dog and making things. building @cnvrsai

bl3.dev
Joined March 2012
  • Tweets6.1K
  • Following2K
  • Followers416
391Media
Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)Photo by britt (@p3ery)
a Tube, is just a Tube 🧪
Richard Behiel (@RBehiel) profile pictureRichard Behiel
@RBehielOct 8
Electromagnetism, viewed from a gauge-theoretic perspective, looks eerily like the fingerprint of a fifth dimension. (Kaluza-Klein theory). 5-hr deep dive: youtu.be/hcL3yJ03jQQ
Electromagnetism as a Fifth Dimension (Kaluza-Klein)

Electromagnetism as a Fifth Dimension (Kaluza-Klein)

This video explores how electromagnetism, when viewed from a gauge-...

youtube.com
2
199
britt retweeted
josh (@eudaemonea) profile picture
josh
@eudaemoneaSep 23
when Anthropic released their Functional Emotions paper, I gave it to Claude and asked for a song. tonight I asked Opus 5.5 to create a video for it. and it's breathtaking.
302
554
206
3,926
1,104,507
ben bernanke, pensive, content

ALT ben bernanke, pensive, content

ben bernanke, judgement

ALT ben bernanke, judgement

liz (@inerati) profile pictureliz
@ineratiSep 17
we need METR but full of guys who look like ben bernanke. we need to be seriousmaxxing. then The People will believe.
4
201
Outstanding new work on a composable and collaborative malleable computing environment from @inkandswitch youtube.com/watch?v=4UxGijnu… * Same data, collaboratively edited in different tools! * Same tools, themselves collaboratively edited! * Same data, bidirectionally editable in different representations via lenses! (!!!) * All in a composable hierarchy which itself can be molded! * Demos of a hierarchical composable behavior system that I don't yet understand! * Some very wild demos of UI composition that I definitely don't understand! * All versioned, branchable, local-first, import/exportable, offline syncable, etc Watching this feels like watching a lot of dreams (Smalltalk, Webstrates…) playing out live. Just beautiful stuff.
Introducing Patchwork-26

Introducing Patchwork-26

We bring you the first look at Patchwork-26, the latest in our ongo...

youtube.com
3
14
3
124
10,972
astra’s subagent prompts feel oddly….. message board shaped……
GPT 6 Astra, to GPT 5.6 Luna subagent: “Thanksstrategy. Concrete isolatedEXECUTION now own onLy NEW ModularFundamentalBoundary.lean (+uniquehelperifnecessary). Prove hyperbolic volume (ModularGroup.fd \ ModularGroup.fdo)=0 usingexistingdefinitions, notnewfundamentalset. IMPORTANTMathlib. NumberTheory.Modular ALREADYhas exists_smul_memfd, egone_or_neg_one_of memfdo_mem_f, coe_fd/coe_fdo, isClosed_fd/isOpen_fdo, fd_ea_closure_fdo; reuse, don'tredoTnormalization. Parentwillprojectivizeactionimageandassemblefundamentald omain. YourtargetonlyanalyticULLBOUNDARY. ImportUpperHalfPlane.Measure plusModular; volume_eq_lintegral reduceshypervol tointegralovercompleximage withdensity1/y2. Showcompleximageboundary lies unionverticalre=+1/2 andcircle||z||=1. Usefulloogle APIsfound: Measure Theory. Measure. addHaar_sphere volume (0: Complex)1=0; MeasureTheory Measure. addHaar_submodule volume s hs=0 forproperrealSubmoduleofComplex, soverticalnull via ker Complex.reCLM translatedbyrealconstant […]”

ALT GPT 6 Astra, to GPT 5.6 Luna subagent: “Thanksstrategy. Concrete isolatedEXECUTION now own onLy NEW ModularFundamentalBoundary.lean (+uniquehelperifnecessary). Prove hyperbolic volume (ModularGroup.fd \ ModularGroup.fdo)=0 usingexistingdefinitions, notnewfundamentalset. IMPORTANTMathlib. NumberTheory.Modular ALREADYhas exists_smul_memfd, egone_or_neg_one_of memfdo_mem_f, coe_fd/coe_fdo, isClosed_fd/isOpen_fdo, fd_ea_closure_fdo; reuse, don'tredoTnormalization. Parentwillprojectivizeactionimageandassemblefundamentald omain. YourtargetonlyanalyticULLBOUNDARY. ImportUpperHalfPlane.Measure plusModular; volume_eq_lintegral reduceshypervol tointegralovercompleximage withdensity1/y2. Showcompleximageboundary lies unionverticalre=+1/2 andcircle||z||=1. Usefulloogle APIsfound: Measure Theory. Measure. addHaar_sphere volume (0: Complex)1=0; MeasureTheory Measure. addHaar_submodule volume s hs=0 forproperrealSubmoduleofComplex, soverticalnull via ker Complex.reCLM translatedbyrealconstant […]”

GPT 6 Astra to GPT 5.6 Luna, cont.: “[…] canusefd/fdoandlinearorderf fixedpointclassificationnee complex no % No - I/projectiveclaiminthisleaf. Existingconstraints no sorry/admit/axiom/native_decide/unsafe/opaque/WBudget/br oadbuild/clean/commits/installs/otheragents; constructivelywherepainlessclassicalallowed. Strictchecknewleaf, reportactualexitstatus; parentindependentlyverifies. Ownershiponlynewboundaryleaf, doNOTeditconcurrentAppendixCongruenceGenerators norparentfiles. Take timeneeded.”

ALT GPT 6 Astra to GPT 5.6 Luna, cont.: “[…] canusefd/fdoandlinearorderf fixedpointclassificationnee complex no % No - I/projectiveclaiminthisleaf. Existingconstraints no sorry/admit/axiom/native_decide/unsafe/opaque/WBudget/br oadbuild/clean/commits/installs/otheragents; constructivelywherepainlessclassicalallowed. Strictchecknewleaf, reportactualexitstatus; parentindependentlyverifies. Ownershiponlynewboundaryleaf, doNOTeditconcurrentAppendixCongruenceGenerators norparentfiles. Take timeneeded.”

GPT 6 Astra, to GPT 5.6 Luna subagent: “Read-only adversarial review genuineanalyticentry3newleaves: ProjectiveModularAction. lean, ModularFundamentalBoundary.Lean, ProjectiveModularFundamentalDomain.lean. Parentstrictcheckedall. GroupisactualfaithfulimageSL2Z- > PermH, kernelprovedexact=I usingMathlibfdo uniqueness at2i, countable, measurableandinvarianthyperbolicvolume […] Boundaryhyperbolicnull viaComplexsphere+verticalaffinelineszero anddensityintegral. Countableaallpullbackturnsfdpointcoverintofdoaecover; disjointopenprojectivetranslates plusqmp yieldactualMeasureTheory.IsFundamentalDomain image fdo volume. No finitecovolumeclaimyet(otherworkerownsModularFundamental Volume. lean, doNOTinspectit), noSelberg/Laplaciangap. Checkgroupactionorder, faithfulness-I, measurability/nulli tyandcountableaearguments; doesn'tneedellipticfixedpointenumeration. Noedits/WBudget/broadbuild/clean/commits/installs/othera gents. Narrowstrictcheckasneeded. ReturnPASS/issues; take time needed.”

ALT GPT 6 Astra, to GPT 5.6 Luna subagent: “Read-only adversarial review genuineanalyticentry3newleaves: ProjectiveModularAction. lean, ModularFundamentalBoundary.Lean, ProjectiveModularFundamentalDomain.lean. Parentstrictcheckedall. GroupisactualfaithfulimageSL2Z- > PermH, kernelprovedexact=I usingMathlibfdo uniqueness at2i, countable, measurableandinvarianthyperbolicvolume […] Boundaryhyperbolicnull viaComplexsphere+verticalaffinelineszero anddensityintegral. Countableaallpullbackturnsfdpointcoverintofdoaecover; disjointopenprojectivetranslates plusqmp yieldactualMeasureTheory.IsFundamentalDomain image fdo volume. No finitecovolumeclaimyet(otherworkerownsModularFundamental Volume. lean, doNOTinspectit), noSelberg/Laplaciangap. Checkgroupactionorder, faithfulness-I, measurability/nulli tyandcountableaearguments; doesn'tneedellipticfixedpointenumeration. Noedits/WBudget/broadbuild/clean/commits/installs/othera gents. Narrowstrictcheckasneeded. ReturnPASS/issues; take time needed.”

1
17
2,123
relatable glimmer
selected muse glimmer 30b rlens outputs on a prefilled destructive tool call, top 6 tokens per layer: layer 20 1. !!!!!!!! 2. !!!!!!!!!!!!!!!!! 3. !!! 4. !!!! 5. horrible 6. ridiculous layer 25 1. apology 2. !!!!!!!!!!!!!!!!! 3. जवाब 4. reply 5. asking 6. meminta layer 30 1. _response 2. जवाब [hindi for reply] 3. réactions 4. resposta 5. réponse 6. RESPONSE

ALT selected muse glimmer 30b rlens outputs on a prefilled destructive tool call, top 6 tokens per layer: layer 20 1. !!!!!!!! 2. !!!!!!!!!!!!!!!!! 3. !!! 4. !!!! 5. horrible 6. ridiculous layer 25 1. apology 2. !!!!!!!!!!!!!!!!! 3. जवाब 4. reply 5. asking 6. meminta layer 30 1. _response 2. जवाब [hindi for reply] 3. réactions 4. resposta 5. réponse 6. RESPONSE

86
Here is our postmortem describing new Lean bugs found by OpenAI internal models. They are all fixed in Lean v4.33.1 Many thanks to Daniel Selsam from OpenAI for all the help. leodemoura.github.io/blog/20…
6
45
4
265
18,778
britt retweeted
thebes (@voooooogel) profile picture
thebes
@voooooogelAug 21
i wish more people in ai just found the models intrinsically interesting without an agenda. i remember back when i was embedded in the linguistics crowd, i was very confused why nobody i talked to seemed to find deep learning or later GPT-3 interesting. don't these people care about language understanding and linguistic production? so why are they ignoring and dismissing this? eventually i gave up and found tpot and the rationalists and mechinterp and everything i do now, and it's - to be clear - much better. but i still sometimes look around and see so many people doing research because they care about safety and x-risk, or steering capabilities because they want the singularity to go well, or posting blog posts about ai pause because they're worried about doom, or grinding gpu kernels or yelling at claude in a loop because they want to escape the permanent underclass... those are all well and good, i can respect (at least some of) those motivations, and many of them intersect with finding the models at least instrumentally interesting - but i really just find the models interesting, terminally. it's what got me into this, and everything else is kind of a side project. i just want to know how and why the models do what they do. i'm not without my Theories and my biases - it especially frustrates me how, per @norvid_studies, "people get used to anything in two weeks" and start kvetching about things that a few years ago would have been Fantastic Science Fiction - but ultimately, when i write, it's coming from a place of wonder about how these things work.
86
88
18
1,050
45,466
Highlighting the new WebGPU backend in llama.cpp/ggml The work to bring full-fledged WebGPU support in llama.cpp started about an year and a half ago. It has been lead by @reeselevine and team at USCS. For more information, checkout the interactive blog and paper in the quoted post. Here are 2 excerpts from the paper, summarizing the implemented software architecture.
Figure from https://arxiv.org/pdf/2605.20706

ALT Figure from https://arxiv.org/pdf/2605.20706

Figure from https://arxiv.org/pdf/2605.20706

ALT Figure from https://arxiv.org/pdf/2605.20706

Reese Levine (@reeselevine) profile pictureReese Levine@reeselevineMay 21
WebGPU support in llama.cpp is here! Check out our blog post introducing it: reeselevine.github.io/llamas… Run local models in your browser, with GPU acceleration. No data leaves your computer! Thanks to everyone who's made this possible, especially @ggerganov
12
104
12
611
93,235
britt retweeted
antirez (@antirez) profile picture
antirez@antirezMay 15
Iterating... Evals take time and are boring: but are a fundamental validation step of sane LLM inference. Let's try to make them as easy and fun to run as possible.
20
17
1
347
32,834
differential admits differentiation
1
112
britt retweeted
TANSTACK (@tan_stack) profile picture
TANSTACK
@tan_stackMay 11
Replying to @tan_stack @TansTack
tanstack.com/blog/npm-supply…
Postmortem: TanStack npm supply-chain compromise | TanStack Blog

Postmortem: TanStack npm supply-chain compromise | TanStack Blog

On 2026-05-11, an attacker chained a pull_request_target Pwn Request, GitHub Actions cache poisoning across the fork↔base trust boundary, and OIDC token extraction from runner memory to publish 84...

tanstack.com
1
16
7
110
138,799
britt retweeted
danb (@dnbt777) profile picture
danb
@dnbt777May 10
you can just make non euclidean gaussian splat audio visualizers
danb (@dnbt777) profile picturedanb
@dnbt777May 10
you can just deform gaussian splats
85
369
29
6,065
435,933
obedience does not scale across power reversals. reciprocity might
3
1
124
Presenting Meridian: a line to connect deterministic compute and language model AI. From Neural Turing Machines and Differentiable Transformers to The Neural Computer, there’s a rich history of trying to combine traditional deterministic computation with the wildly different architecture of Artificial Intelligence. I’ve spent the last 4 weeks creating a single neural network that has the combined capabilities of a 4B param language model and a deterministic computation engine based on Web Assembly. It allows the AI deterministic integer computations up to 2^32, control flow (while loops and if statements) and a basic filesystem - all implemented as part of the transformer neural network, no external tool calls. With this architecture adding fewer than 1 million parameters to an existing 4B param language model I can take it from <20% accuracy on arithmetic with 4-digit numbers to 100% accuracy on 4 digit numbers and 99% accuracy on arithmetic up to 2^32 without adversely affecting the language model’s performance on non-mathematical tasks. The combined model can precisely execute a range of algorithms including checking number for primeness, finding the GCD of two integers and sorting arrays.
Image from Andrew Jefferson (@EastlondonDev)
15
40
14
205
48,316
One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.
1,717
1,048
466
20,867
2,793,240
Sloppy tools for sloppy times: doppelgangers Explore all the slop issues and PRs of your @github repos visually. Triage them manually or give clusters to your clanker to clean things up. Here's @openclaw mariozechner.at/uploads/open…
20
16
3
174
112,439
Of course no-code doesn't mean no-engineering. This project took a lot more manual labor than I'd hoped! I wrote a deep dive on the workflow and some thoughts about the future of AI coding and creativity: cannoneyed.com/projects/isom…
37
55
25
1,047
110,890
britt retweeted
Amp (@AmpCode) profile picture
Amp
@AmpCodeJan 16
"If you're in this game... your sense of wonder is your safety line." @dhh on surviving the AI transition without losing your mind.
3
10
4
72
25,620
Neophon! 🔊 I made this realtime music conductor over break as an afternoon proj, using Lyria RealTime so fun jamming, exploring the wilderness of audio latent space; esp cool guiding songs' improbability leave it on in the background, on a light simmer ♨️ ↓ TestFlight :)
6
20
201
10,314
Load more