iiinova — where your digital self keeps evolving

Reposted by Deyan Ginev

arXiv has updated our policy on rate limiting for all submitters. This update was made to fairly distribute moderator time & support the arXiv community of staff, volunteers, readers & authors. Please read our announcement to learn more: blog.arxiv.org/2026/10/01/updated-r…

110043

Reposted by Deyan Ginev

Robin Berjon @robin.berjon.com · 30/09/2026

The social internet — for science! I'm super excited to see @aster.id launching. It's an Atmosphere service provider (like @eurosky.social) for researchers, hosted by your friends at the @modalfoundation.eurosky.social. Why does it matter? Thread time! 🧵

16924

Reposted by Deyan Ginev

mr. TIM @timkellogg.me · 28/09/2026

Claude 5.5 Sonnet is live and it’s roughly Opus 5.5 but cheaper www.anthropic.com/claude-sonne...

A benchmark comparison table titled "Claude Sonnet 5.5" comparing four models: Sonnet 5.5, Sonnet 5, Opus 5.5, and GPT-6 Sol.
* Agentic coding (Terminal-Bench 4.0): Sonnet 5.5: 70.6%; Sonnet 5: 10.3%; Opus 5.5: 66.4%; GPT-6 Sol: N/A.
* Agentic coding (FrontierCode 1.1 Main): Sonnet 5.5: 46.2% (Max) / 52.1% (Xhigh); Sonnet 5: 42.4%; Opus 5.5: 54.4%; GPT-6 Sol: 49.3%.
* Agentic coding (CursorBench 4.0): Sonnet 5.5: 55.5%; Sonnet 5: 34.1%; Opus 5.5: 57.8%; GPT-6 Sol: N/A.
* Knowledge work (GDPval-AA v2.1): Sonnet 5.5: 1844; Sonnet 5: 1449; Opus 5.5: 1846; GPT-6 Sol: 1487.
* Knowledge work (AA-Briefcase v1.1): Sonnet 5.5: 1811; Sonnet 5: 1359; Opus 5.5: 1822; GPT-6 Sol: 1483.
* Multidisciplinary reasoning (Humanity's Last Exam with tools): Sonnet 5.5: 64.5%; Sonnet 5: 54.9%; Opus 5.5: 67.7%; GPT-6 Sol: N/A.
* Computer use (OSWorld 2.1 partial): Sonnet 5.5: 80.1%; Sonnet 5: 57.0%; Opus 5.5: 81.8%; GPT-6 Sol: N/A.
* Visual chart recognition (Chartography no tools): Sonnet 5.5: 61.6%; Sonnet 5: 15.6%; Opus 5.5: 64.4%; GPT-6 Sol: 53.6%.
Footnotes provide methodology notes regarding evaluation settings, Artificial Analysis pre-release testing details, and recent bug fixes affecting GPT-6 Sol benchmark scores.

A line graph titled "Agentic coding by effort level" on the CursorBench 4.0 benchmark, plotting Score (%) on the linear y-axis (20% to 60%) against Cost per task in USD on a logarithmic x-axis ($0.5 to $10+).
The chart compares four models:
* Sonnet 5.5 (blue line with labeled effort levels): Starts at "Low" (~$0.50, 36%), moving through "Med" ($0.70, 39%), "High" ($1.70, 48%), "Xhigh" ($3.80, 53%), and "Max" ($9.50, ~55.5%).
* Opus 5.5 (orange line): Tracks closely above Sonnet 5.5 at higher cost points, spanning from ~$1.20 per task (~44%) up to ~$12.50 per task (~58%).
* Sonnet 5 (grey line): Shows lower accuracy relative to cost, ranging from ~$0.90 per task (~25%) to ~$8.00 per task (~42%).
* GPT-5.6 Sol (light green line): Represents the lowest trajectory, ranging from ~$1.40 per task (~24%) to ~$7.00 per task (~34%).
Footnote: "CursorBench 4.0 does not report GPT-6 Sol performance publicly, so we report GPT-5.6 Sol here."

2020623

Deyan Ginev @dginev.bsky.social · 26/09/2026

Opus 5.5 is extremely efficient, solves problems with little revision, yet I still used the weekly quota reset 4 days in. We are not running out of software to write any time soon ...

000

Deyan Ginev @dginev.bsky.social · 23/09/2026

Keep rooting for arXiv! We are in exciting times for Open Science.

090

Reposted by Deyan Ginev

Curvenote® @curvenote.com · 14/09/2026

We ship in the open. Our Changelog is public and written for the people actually using Curvenote products - new features explained, UX improvements and product updates. If you have ever wondered what we are up to go check us out curvenote.com/changelog

curvenote.com

Changelog | Curvenote

The latest updates and improvements from Curvenote.

051

Reposted by Deyan Ginev

Robin Berjon @robin.berjon.com · 07/09/2026

Hey internet people, is inline LaTeX something you'd be interested in? #experimunt

3315120

Deyan Ginev @dginev.bsky.social · 07/09/2026

Great retrospective. This sentiment by John MacFarlane also rings true for LaTeXML and all markup converters: "Indeed, a time may come when LLMs can produce more reliable conversions than pandoc or anything that works like it." pandoc.org/twenty-years...

110

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 04/09/2026

As it becomes clear that language models—like humans—love passing notes to each other on message boards, I’m starting to think the thing we need to worry about is not the “alignment” of an isolated model but, like, the fluid dynamics of a language slurry.

1630831

Reposted by Deyan Ginev

Lea Verou, PhD @lea.verou.me · 04/09/2026

Came across this fantastic post by Slava Shestopalov randomly while looking for something else medium.com/design-bridg... In the usability & web tech class we used to teach at MIT with @karger.bsky.social I’d give a lecture on graphic design that included this, but this is so much better presented.

medium.com

Optical effects in user interfaces

How to make optically balanced icons, correct shapes alignment, and perfect corner rounding

14512

Deyan Ginev @dginev.bsky.social · 03/09/2026

I am surprised I am still this surprised

010

Reposted by Deyan Ginev

Carl T. Bergstrom @carlbergstrom.com · 03/09/2026

"Why is it so hard to find peer reviewers?", asks guy who has been publishing 15 papers a year for the last decade.

317430

Reposted by Deyan Ginev

Lisa DeBruine @debruine.bsky.social · 03/09/2026

Wow. This is incredibly convincing evidence of falsification with clear intent. Huge props to Klaus Wertenbroch, who is probably having a pretty stressful time.

0244

Reposted by Deyan Ginev

Grant @syncingsinking.bsky.social · 01/07/2026

Roses are hot Violets are broader

Tue, 30 Jun 2026 (showing 15 of 15 entries)

[1] arXiv:2606.30422 [pdf, html, other]

The Conway knot has infinite concordance order

Chiara Donatone, Marc Kegel, Lukas Lewark, Paula Truöl

Comments: 14 pages, 4 figures, comments welcome

Subjects: Geometric Topology (math.GT)

011

Reposted by Deyan Ginev

mr. TIM @timkellogg.me · 26/08/2026

it’s official: GLM-5.3-Flash is Ox Alpha! z.ai/blog/glm-5.3...

Line graph titled "Agentic Coding Performance by Effort Level" plotting accuracy percentage against average output tokens per task for five AI models on Z.ai Code Bench v1.0:
* Claude Fable 5: Highest performance overall, scaling from ~28.8% accuracy at 33K tokens (Low effort) to ~39.4% accuracy at 114K tokens (Max effort).
* GLM-5.3: Second highest performance, rising sharply from ~24.6% (Low effort) to ~31.3% (High effort) near 49K tokens, peaking at ~34.5% accuracy at 74K tokens (Max effort).
* GLM-5.3-Flash: Scales from ~21.5% accuracy at 41K tokens (Low effort) to ~29.0% accuracy at 139K tokens (Max effort).
* Claude Opus 4.8: Scales from ~21.2% accuracy at 28K tokens (Low effort) to ~29.5% accuracy at 120K tokens (Max effort).
* GLM-5.2: Lowest overall performance, moving from ~19.4% accuracy (Non-Thinking effort at 42K tokens) to ~23.4% accuracy at 94K tokens (Max effort).

711712

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 18/08/2026

We should cut each other a lot of slack on the AI front, tbh, because what's happening is way outside our past experience, and no one has very good intuition for what's probable or what the implications are.

511313

Deyan Ginev @dginev.bsky.social · 15/08/2026

The knights who say <mi>ni</mi>

130

Deyan Ginev @dginev.bsky.social · 15/08/2026

Nice idea - web math is also RSS math.

010

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 14/08/2026

"If the models need a model, they can model their answers on ..." => the word "template" about to have its best decade ever

3213

Deyan Ginev @dginev.bsky.social · 12/08/2026

Today we add fresh arXiv HTML for 98.8% of articles with LaTeX sources, available immediately. Full summer reset. Rust rewrites rock!

0203

Deyan Ginev @dginev.bsky.social · 10/08/2026

Slowly at first, then all at once.

020

Reposted by Deyan Ginev

Philip Ball @philipcball.bsky.social · 10/08/2026

"Restructuring research funding as a choice between discovery and application gets the economics wrong, the evidence wrong and the people wrong." Great piece. I suspect Richard Jones (on Bsky?) will have views. arxiv.org/abs/2608.00172

arxiv.org

Discovery is not the opposite of application

Restructuring research funding as a choice between discovery and application gets the economics wrong, the evidence wrong and the people wrong.

33015

Deyan Ginev @dginev.bsky.social · 05/08/2026

🗓️ The July 2026 arXiv articles are now in ar5iv. The Rust conversion is really making a difference, these updates are now much faster/cheaper to complete. Some recent samples if curious: ar5iv.labs.arxiv.org/html/2607.12... ar5iv.labs.arxiv.org/html/2607.21... ar5iv.labs.arxiv.org/html/2607.25...

ar5iv.labs.arxiv.org

The homotopical monadicity theorem

We give a broad homotopical analog of the classical categorical Beck monadicity theorem. It often holds when classical monadicity fails. This grew out of an understanding of a general context for reco...

030

Reposted by Deyan Ginev

Alyssa Smith @cetaceanneeded.bsky.social · 04/08/2026

I'm incredibly honored to be one of the inaugural winners of the Ctrl-Z award. The sequence of events preceding and following this retraction were harrowing (and are honestly still stressful), but I hope this will be an opportunity for more open dialogue about mistakes + self-retraction in science.

513836

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 31/07/2026

Ender’s Game, except we’re the aliens

311214

Reposted by Deyan Ginev

Richard Sever @richardsever.bsky.social · 30/07/2026

arXiv has spun out as an independent non-profit like openRxiv. Today they announce their new CEO and founding Board blog.arxiv.org/2026/07/30/a...

blog.arxiv.org

arXiv welcomes inaugural CEO and founding Board of Directors

05131

Reposted by Deyan Ginev

New-Cleckit Dominie @ncdominie.bsky.social · 28/07/2026

This is just to say I have OCRed the text thet was in tne arch:ve and whicb yov were prabaldy seamhing for "breahtast" Fargme ms fheg mem debo:w5 sb ?mccf mmb 5c oc1b

7778512571

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 27/07/2026

There’s resonance between Timothy Gowers’ reasons for not signing the Leiden Declaration, and @rawx.bsky.social ‘s post on “Interpolatable Archives,” which connects Gowers’ vision of a “mathematical culture” beyond “discovery” to related theses by Borges, Shalizi & Farrell.

open.substack.com

Explosion Drawings

On Interpolatable Archives, Part 2: Science Sans Discoveries, Textrotating Cognitive Catalysts and Exploding Your Intelligence by the Method of Warburg

3132

Reposted by Deyan Ginev

openRxiv @openrxiv.bsky.social · 23/07/2026

The Curvenote Reader button is live on bioRxiv! 😎 Click it to explore papers in a new way—zoom figures, pull up references without leaving the page. Button appears 48 hours after papers drop. Feedback? DM us or hit the GitHub repo. openrxivlabs.org/articles/rea...

openrxivlabs.org

Why we’re expanding the openRxiv Labs Curvenote Reader experiment

Read on for early results and what we’ll be looking at in the next phase.

02211

Deyan Ginev @dginev.bsky.social · 17/07/2026

cargo +nightly install latexml

110

Reposted by Deyan Ginev

Deyan Ginev @dginev.bsky.social · 05/07/2026

📢 The June 2026 arXiv articles are now in ar5iv, with three big changes: Rust, NYC, AI 🧵1/5

124

Deyan Ginev @dginev.bsky.social · 05/07/2026

📢 The June 2026 arXiv articles are now in ar5iv, with three big changes: Rust, NYC, AI 🧵1/5

124

Deyan Ginev @dginev.bsky.social · 03/07/2026

🎙️ ... is this thing still on? LaTeXML announcement season starting with the next ar5iv update (Sunday, 5th of July) This time it's proper BIG news. #Rust

130

Reposted by Deyan Ginev

Mike Boylan-Kolchin @mbkplus.bsky.social · 29/06/2026

And it’s another reminder that there’s no place for American researchers to go en masse if science in the US is further degraded. There’s no replacement at scale.

0468

Reposted by Deyan Ginev

Mike Boylan-Kolchin @mbkplus.bsky.social · 29/06/2026

I hate that this — urgent pleas to save prestigious departments from drastic cuts — is becoming a regular occurrence. A program like this takes decades to build and a few strokes of a pen to destroy.

sussexphysics.github.io

Save Sussex Physics

Save Sussex Physics - Open Letter

215069

Reposted by Deyan Ginev

pamelafox.bsky.social @pamelafox.bsky.social · 23/06/2026

I printed a custom t-shirt that's an ode to @simonwillison.net's Pelican benchmark. The super chill highly saturated cruising pelican came from this Gemini 3 launch in Feb: simonwillison.net/2026/Feb/12/... (And the caption was inspired by the reaction on social media to the blog post.)

1796

Reposted by Deyan Ginev

Ryan @snarfed.org · 17/06/2026

no no, you don't understand, it's not a torment nexus it's just a torment fidget spinner

0374

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 13/06/2026

TFW you’re not going to do anything about Jesus of Nazareth — or take Herod’s name off the temple. You just really love building scaffolding.

This Jesus Must Die (From "Jesus Christ Superstar" Soundtrack)
Bob Bingham, Kurt Yaghjian https://m.youtube.com/watch?v=lN4TxOd9I8Q&ra=m

0223

Deyan Ginev @dginev.bsky.social · 06/06/2026

🗓️ The May 2026 arXiv articles are now in ar5iv.

041

Reposted by Deyan Ginev

Ted Underwood @tedunderwood.com · 05/06/2026

I respect the authors and believe them sincere, but ppl who are sincerely concerned about RSI need to start considering ways of diffusing risk by diffusing power and access — not by “internationally coordinated slowdown,” which as Shubhendu says is not happening.

8605

Reposted by Deyan Ginev

Randall Munroe @xkcd.com · 03/06/2026

Sunbeam xkcd.com/3253/

Comic. [Person 1 talking to a person with shoulder-length hair and a person with a white hat, gesturing to two tables with three chairs each, a window above shining light onto the floor in between them.] Let’s take the far table. The closer one will be in the sun soon. [caption] Astrophotography gives you exactly one extremely minor superpower.

204004458

Deyan Ginev @dginev.bsky.social · 02/06/2026

internet: "self-publishing is now possible!" humanity: /wp-content/uploads/2026/06/main.pdf

011

Reposted by Deyan Ginev

openRxiv @openrxiv.bsky.social · 01/06/2026

Introducing openRxiv Labs! 🔬 Growing from the strong foundation of bioRxiv & medRxiv, Labs is an experimental space for working with collaborators to push the boundaries of open scientific communication. 👉 Check it out and learn more: openrxiv.org/openrxiv-lab... #OpenScience #Preprints #Scicomm

openrxiv.org

Launching openRxiv Labs - openRxiv

Over the last thirteen years, bioRxiv and medRxiv have grown into widely used infrastructure for rapid research sharing in biology and medicine. The reliability and researcher-first values that have d...

33727

Reposted by Deyan Ginev

Melanie Mitchell @melaniemitchell.bsky.social · 21/05/2026

"Science is not simply a sequence of tasks that can be optimized....The scientific system thrives on inefficiency: redundant efforts, failed attempts, and divergent paths. These are not costs to be eliminated but sources of discovery." Excellent essay! ⬇️ www.science.org/doi/10.1126/...

science.org

Progression without progress

Progress in the use of artificial intelligence (AI) to advance scientific discovery has made it increasingly realistic to envision automated “end-to-end science” (ETES) systems: integrated pipelines t...

1421150

Deyan Ginev @dginev.bsky.social · 18/05/2026

The tragedy of our time is that we let our attention and intensity go from 0 to 100 on short-term concerns ("arXiv is banning users for a new reason"), ...but can only go from 0 to 5 on long-term concerns ("keep all arXiv articles healthy and freely available in 2226")

120

Reposted by Deyan Ginev

Ana Tudor @anatudor.bsky.social · 14/05/2026

In case you didn't know, if you have a border and the `border-color` isn't set, not even in the browser stylesheet, then the `border-color` defaults to `currentColor`. #CSS

HTML:

```
<button>I'm a button, I have a border in browser stylesheet</button>
<div>I'm a div, I have no border in the browser stylesheet</div>
```

CSS:

```
/* this produces a crimson border only on the div
* since the button already has a border-color set
* in the browser stylesheey */
div, button {
color: crimson;
border-style: solid
}

/* this produces a cimson border on both the div and button
* since the border shorthand overrides the button's
* border-style, border-width and border-color
* from the browser stylesheet and, since it
* only explicitly specifies a border-style,
* border-width defaults to medium and
* border-color defaults to currentColor */
div, button {
color: crimson;
border: solid
}
```

0323

Deyan Ginev @dginev.bsky.social · 06/05/2026

🗓️ The April 2026 arXiv articles are now in ar5iv.

010

Deyan Ginev @dginev.bsky.social · 04/05/2026

I have 9% claude weekly quota left for 2 workdays. Time to read those "token budget" blog posts that kept flying by. Send me your best advice.

000