iiinova — where your digital self keeps evolving

Philipp Markolin @philippmarkolin.eurosky.social · 22h

I get to this age now where I go to get a coffee and snack for myself, instead of rushing to do the next thing.

Philipp Markolin @philippmarkolin.eurosky.social · 15/10/2026

What´s old is new again. Reminds me of Nicholson Baker; the fiction writer who landed his "The lab leak hypothesis" in NYT Magazine in January 2021, and it became the most popular story of the year. This is what he believes about various diseases though ⬇️

Nicholson Baker was, of course, not new to the cover-up conspiracy genre, and the idea of laboratory accidents especially seems to have enamored the writer for years. For example, in his book Baseless, a story woven around his frustration with governmental redactions of 1950-era bioweapon research documents, he casually asserts that “rabbit fever, Q fever, bird flu, Lyme disease, wheat stem rust, African swine fever, and hog cholera all look, to my nonscientist’s eye, like unnatural epidemics that owe their outbreaks to the laboratory.” He felt a deep unease about biotechnology and had the literary skills to put his worries into words.

2101 10

Philipp Markolin @philippmarkolin.eurosky.social · 10/10/2026

Remember when the commentators were telling you that Jay is a great choice and save pair of hands for restoring trust in the NIH? Guess they are on board with his new RFKjr's "vaccine injury" health clinic? Trust restored indeed 🫠

161 6

Philipp Markolin @philippmarkolin.eurosky.social · 08/10/2026

On OpenAI's solution to Navier-Stokes in lean, some confusion remains one the natural language version. What if the NL paper declares stronger statements than what Lean actually proves? "Developing semantically faithful AI autoformalisation stands as a major open problem" arxiv.org/pdf/2610.08144

ABSTRACT. Autoformalisation is increasingly used to verify mathematical texts, including those generated by AI,
as in OpenAI’s announced proof of blow-up of solutions to the Navier-Stokes equations. In this process, an AI
system translates the text from a natural language (NL) into a formal language such as Lean. Once this translation
is done, the argument expressed in the formal language can easily be mechanically verified. The purpose of this
article is to demonstrate why this process may offer no confidence in the original NL argument, owing to the vari-
ous difficulties in performing the translation semantically faithfully. In particular, we highlight that the problem of
resolving ambiguities in mathematical NL text, which is necessary in order to provide semantically faithful trans-
lation, is arbitrarily high up in the Solvability Complexity Index (SCI) hierarchy/arithmetical hierarchy (the SCI
= ∞). Hence, informally, providing semantically faithful AI autoformalisation is harder than any computational
problem including the Halting problem (which has SCI = 1). To demonstrate the effect of this result we provide
several examples of AI mistranslations of NL statements and proofs into Lean in practice, resulting in mismatches
between NL proofs and their Lean ‘verifications’. These include OpenAI’s announced Navier-Stokes proof. In
particular, we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to
the Navier-Stokes equations.

150 5

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

So now is not the time to despair or give in oligarchic hyperscalers, fascist movements and their dystopian visions of our future. Now is the time to do. To experiment. To build. And to fight for our pluralistic future. To birth the new world that will be better than the old world that is dying.

030 3

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

But that is just a start, and civic engineers like @audreyt.org have been showing (and initiating) many visions on how pro-social engineering can look like, from bridge-building algorithms to AI as kami or community-based "AI in the loop of humans"; a horizontal, not vertical integration of AI.

130 3

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Lastly, to really protect democratic pluralism, we need to participate, to build, to engineer and to support the pro-social systems we want to have. Luckily, many such initiatives are happening, including in and on @bsky.app itself, such as the @mysky.social feed from @noupside.bsky.social.

130 3

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

... would fall into that sector. This includes also regulations such as the European digital services act and work from academics like (my go-to) @lewan.uk who really think hard about what it means to preserve human rights, dignity and deliberation online We also need defensive cyber capabilities

130 3

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

On the right, we have the Habermasian idea of "deliberative" democracy, because voting does not mean much if people are so manipulated by propaganda they are not free to make their own choices; so anything we can do on data privacy, digitial rights, choice architectures and defensive capabilities...

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

enables the power of the US tech hyperscalers to dominante the EU and the wider world. As @doctorow.pluralistic.net argues, if Microsoft were to shut down in EU tomorrow, total mayhem would ensue and billion dollar businesses would be effectively shut down. Not a great place to be in.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

... on particular interrelated problems of our current digital infosphere woes. Left, I grouped everything that has to do with technological representation; from digital sovereinty to neutral infrastructure to ownership and regulation of tech. Vertical integrated, captured infrastructure is what..

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

That is how I left the students, but I realized that many, if not all of them, have no idea where to even start. So I thought I share my - very much developing understanding - of what solution spaces look like, here as well, maybe with some comments. This is a map based on many experts working...

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

So now in this period where things are changing, we have the opportunity to shape the future we want to live in. We need to make use of our agency and understanding on what is going wrong today to not fall into the same problems. For that: We need to act. We need to experiment. We need to build.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Now to spur discussions, I highlighted some recent analyses and different visions from @newpublic.org (bit.ly/4hzzuLX) and @pettertornberg.com (bit.ly/4hLaT5w) on how AI is changing media economics again. I think the old world is dying, and the new one not yet born.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Now while this is just one framing to understand the current moment, the fragmentation of reality perceptions over the last 15 years has created real problems for discourse and democracy that are not easy to solve. Into this, we throw the next disruptive media technology with AI.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Because you need shared facts to have a basis for discussion upon which solutions to problems can be democratically found. If up is down, and blue is red between different communities, no consensus can be found and factional warfare ensures; the need to dominate and excert power over others.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Market asymmetries and media manipulators basically drive facts and shared reality perceptions from our shared discourse spaces, instead creating a mirror world of engaging fictions that most of us passively consume, participate in, and build our worldviews around. Bad for democratic consensus.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Next, I introduce a class of professional media manipulators who long learned to work the attention economy in their (or their sponsors) favor, using various tactics and strategies of marketeers, propagandists and information combatants to further their financial, strategic and ideological goals.

120 2

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

In an asymmetric playing field, the "best" or most factual information does not rise to the top or get a fair hearing. There is no market mechanism for it. Instead, the information that best captures attention does, and you can optimize information products much better if you dont care about facts

131 3

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

We moved from information scarcity to information abundance. But information abundance shifted the market to a new scarcity: Attention. Hence the attention economy. But attention is not evenly distributed, it follows a power law. Hence information products dont find an even playing field.

110 1

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Was invited to give a lecture for media economy bachelor students today. I am afraid I feft them with a bit of an apocalyptic sense on where we are today. 😅 So I just wanted to outline a little bit my thinking and why I believe now is the time to build something better; and that gives me hope.

1104 10

Philipp Markolin @philippmarkolin.eurosky.social · 07/10/2026

Do people still post touristic pics here? Seems like I do.

060 6

Philipp Markolin @philippmarkolin.eurosky.social · 03/10/2026

🥊 Season 2 of #ScienceCounterpunch is here! We open with biosecurity expert Dr. Gigi Kwik Gronvall, who just testified before Congress on the abysmal state of US biopreparedness. All while H5N1 bird flu has hit US farmed mink for the first time recently. Are we ready? Short answer: fuck. 🧵👇

1169 16

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

After a little playing around with their tool, I can kind of produce some of what we see in the claude opus video. www.mysimulator.uk/math/phyllot...

150 5

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

After that, the only question I had left is why claude Opus 5.5 seemed so much better than the other AI slop; specifically some of the "run simulations" were rather impressive. Turns out that is probably a feature of this free-use browser interactive tool. www.mysimulator.uk/math/phyllot...

141 4

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

Now one question I still had is why from all the niche science topics in the world, all those slop videos would pick this rather obscure topic when you ask it "pick a problem" Turns out, theyoutube channel numberfile did a video about it in 2018 that got ~4 mil views www.youtube.com/watch?v=sj8S...

172 7

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

... is the publication arm of a non-profit that has been running since at least 2007 with really high-quality production of articles. They also have a lot of science writer contributors, illustrators, photo editors etc. Certainly a gem and team effort here. asknature.org/about/

170 7

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

All these short slop videos are variations of each other, but still in a somewhat iterative distinct visual style. This suggests they are sourced probably from a single source, rather than re-created from text. So where did the rather specific visual content come? I think I have an idea 🔽

https://asknature.org/strategy/fibonacci-sequence-optimizes-packing/#related-content

181 8

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

First thought was of course to check whether this creative direction is maybe something it was taken from a human creator. So I checked on youtube, and it turns out that there is a whole slop factory of almost identical sunflower/fibbonacci content, spanning months back. Similar styles & words.

171 7

Philipp Markolin @philippmarkolin.eurosky.social · 26/09/2026

Random little investigation I saw that Claude Opus 5.5 can now create quite appealing "science explainer" shorts as one-shot prompt: "Pick a problem or mystery that obsesses you and solve it as best you can & make a movie we can share on social media about it" So I wondered where info came from?

31716 17

Philipp Markolin @philippmarkolin.eurosky.social · 20/09/2026

What adds a nice touch is that when you start playing with the levers, you can directly observe by how your feed changed; For example, which posts moved up the rank, or which new ones came in.

020 2

Philipp Markolin @philippmarkolin.eurosky.social · 20/09/2026

The algorithmic backbone / infrastructure is fairly basic, what is interesting or more novel is to give agency to users in a transparent way. If you want to tune your algorithm, you basically log into here: app.greenearth.social#/settings/yo... You will find a simple API to tune those levers.

121 2

Philipp Markolin @philippmarkolin.eurosky.social · 10/09/2026

Worse for OpenAI, the sophicity conjecture they solved using the same approach as Andreas Thom was the only one of consequence that actually enabled their hype. Of course we dont know if any of the other problems were truly solved independently or stolen as well. techjournal.org/openai-astra...

What Is the Non-Sofic Group Result, and Why Does It Stand Out?

The non-sofic group construction is widely considered the most significant of the 10 results. The headline result is the construction of the first known non-sofic group, a counterexample to a question that stood open since Mikhail Gromov introduced the concept of soficity in 1999. A group is sofic if its behavior can be approximated arbitrarily well by finite permutation systems; all amenable groups and all residually finite groups satisfy this condition. Whether every countable discrete group must be sofic was one of modern group theory’s most prominent unanswered questions for 27 years.

Astra’s construction answers it: no, not every group is sofic. The implications extend to operator algebra theory, ergodic theory, and the study of cellular automata. University of Manchester mathematician Thomas Bloom, who curates the erdosproblems.com database, weighed in publicly. Bloom called the results “big news” in an X post, adding that, as mathematical constructions, they were bigger than the unit-distance counterexample OpenAI had disclosed earlier.

The second most-discussed result concerns a foundational problem in algebra. The second major result is a disproof of the Connes Rigidity Conjecture, posed by Fields Medalist Alain Connes in 1980. The conjecture concerns whether the von Neumann algebra L(G) constructed from a property (T) group G retains enough information to reconstruct G. Astra disproved this conjecture by constructing infinitely many non-isomorphic groups with property (T) that share the same von Neumann algebra, and it also proved Ehrhart’s volume conjecture.

1121 12

Philipp Markolin @philippmarkolin.eurosky.social · 10/09/2026

He says that he "opted out on June 29." that any of his data would be used for training the model.

3/3 Sellke’s categorical answer was, at minimum, unjustifiably broad and materially misleading; looking back it was plainly dishonest. Bubeck’s acknowledged attacks and his attempt to exclude Alpöge from a proposed OpenAI-proof paper deepen the concern that there is a loss of moral compass. Taken together, this raises serious questions about their judgment and personal integrity.

1100 10

Philipp Markolin @philippmarkolin.eurosky.social · 10/09/2026

OpenAI used their creative path to solve the conjecture after the mathematicians had used ChatGPT. "There is a certain (frankly unacceptable) lack of transparency here; and I fear it will damage the communal process of math more than the new AI-generated results will benefit the subject."

2/3 OpenAI is drawing a distinction that its answer to me erased, despite the fact that my question explicitly made that distinction.
We are not required to reverse-engineer OpenAI’s internal training pipeline to establish what happened. Only OpenAI has the relevant data for that. If Sellke and Bubeck give a categorical denial, it must disclose the basis: product and privacy settings, relevant datasets and checkpoints, and what “de-identified data derived from usage” means.
I disabled model training on 29 June. That control is still only a promise whose implementation users cannot audit, and it is prospective: it does not answer what happened to earlier conversations or to derivatives already selected. OpenAI’s answer did not mention the setting or any account-specific check.
There is also a mathematical parallel, though a less explicit one. The approach of Kun and myself was not the main line of attack on non-soficity, in fact there were other more promising approaches along the line of quantum games etc. OpenAI’s detailed command of our techniques therefore made me wonder how the model found this route. If nonpublic research supplied by users improved a model and the provider then used that model to race those users to publication - without informed consent, disclosure, or credit - that would be ethically indefensible. De-identification may remove a name; it does not remove the intellectual content of a mathematical idea. Sellke and Bubeck seem to be blind to this simple moral aspect.

2121 12

Philipp Markolin @philippmarkolin.eurosky.social · 10/09/2026

🚨🚨🚨 Seems like another Mathematician comes forward with concerns about OpenAI stealing unpublished math research and then publishing it as AI solve under big publicity. Andreas Thom and Gábor Kun worked on Gromov’s soficity conjecture. (See his thread with alt-text) mstdn.science/@andreasthom...

1/3 I want to record a number of points that I realized after reading up on the controversy. In fact it recalls an exchange I had with OpenAI after its non-sofic-group announcement. It exposes a serious, unanswered question about whether researchers can trust OpenAI with unpublished mathematics.

I wrote in an email to Mark Sellke and Sebastien Bubeck shortly after the surprising finding of a non-sofic group using the methods of Kun and myself:
“Another point is that I and a colleague in Dresden were discussing the expander matching problem and various extensions of the work with Gabor Kun actively over the last months with ChatGPT, so that we are of course curious if that was part of the training data or accessible to the reasoning process. There is a certain (frankly unacceptable) lack of transparency here; and I fear it will damage the communal process of math more than the new AI-generated results will benefit the subject.”

Mark Sellke’s complete answer was: “Regarding your conversations with ChatGPT: that did not happen.”

I had explicitly asked about two different things: (1) whether our conversations entered training data, and (2) whether they were accessible to the solving process. The categorical answer now looks as though it addressed only direct access under (2). No such qualification, explanation, or evidence was given. I take this as dishonesty to say the least.
OpenAI now says in the Buckmaster-Alpöge case that no specific user data was accessed, but adds that it “cannot rule out that de-identified data derived from their usage of our products helped improve our models.”

55536 55

Philipp Markolin @philippmarkolin.eurosky.social · 09/09/2026

The good old: "If you go public, just be careful not to destroy your career and btw, we will be done playing nice" Original statement from Tristan (NYU Mathematician): cims.nyu.edu/~tristanb/st...

2113 11

Philipp Markolin @philippmarkolin.eurosky.social · 09/09/2026

The second point of suspicion here is of course the marketing. Oh, they will not claim the prize because that does not matter (and might, down the line, cause them legal trouble if the espionage is proven). It's just about AI being so powerful, so you better keep investing.

171 7

Philipp Markolin @philippmarkolin.eurosky.social · 09/09/2026

So here are a few questions. A) Why are they not specific about what happened and when? B) Why are they suspiciously "generous" to the mathematicians, if according to them, they solved it before them? C) Why are their arguments about proof difference so weak? I think an investigation is needed

2123 12

Philipp Markolin @philippmarkolin.eurosky.social · 09/09/2026

So happy you enjoyed it! One of the starkest contrasts in the book is where you have Alice Hughes, an extraordinary reseacher creating critical origin data under real oppression and danger; and on the other side, you have Alina Chan play-pretend being a suppressed whistelblower in major outlets. 🙄

050 5

Philipp Markolin @philippmarkolin.eurosky.social · 08/09/2026

I am sure the Coal Baron is magically gonna reverse his position on lab leak when a Sarbecovirus with a functional FCS at S1/S2 is found, right? Of course not, he will double down and claim that when nature created it, isn´t it so much more likely the lab found it secretly and used it as template?

2155 15

Philipp Markolin @philippmarkolin.eurosky.social · 04/09/2026

Maybe higher ed blocked you and hid your reply? For me seems fine

100 1

Philipp Markolin @philippmarkolin.eurosky.social · 04/09/2026

And this is probably the most elegantly written concluding sentence I have come across this year. *chef's kiss*

192 9

Philipp Markolin @philippmarkolin.eurosky.social · 03/09/2026

So the "US is hiding 12 alien spacecraft" guy is calling somebody talking about climate-change related disasters a grifter? Is it just me to imagine how the world could be a much better place without all those shameless liars and billionaire propagandists like Shellenberger getting all attention?

2548 54

Philipp Markolin @philippmarkolin.eurosky.social · 02/09/2026

@ostprog.bsky.social Radio Corax kriegt nen shout-out bei Simplicissimus!!!

110 1

Philipp Markolin @philippmarkolin.eurosky.social · 31/08/2026

Seems like conspiracy theorist Bryce Nickels is using his position at Rutgers, a taxpayer-funded university, to parade around a who-is-who of lab leak conspiracy theorists. Challenging Stanford for crank city central I see. 😂 sciencefromthefringe.substack.com/p/announcing...

4258 25

Philipp Markolin @philippmarkolin.eurosky.social · 29/08/2026

You need to get the black and white profile shot ...

030 3

Philipp Markolin @philippmarkolin.eurosky.social · 29/08/2026

Tbf, conspiracy theories are the new lifestyle reporting. Maybe Ornstein wants to change propublica's direction?

070 7

Philipp Markolin @philippmarkolin.eurosky.social · 29/08/2026

Fair point. Cue the next 3 op-eds about how his feelings got hurt by the woke mob, that's why he is endorsing the fascists now.

1213 21

Philipp Markolin @philippmarkolin.eurosky.social · 29/08/2026

Maybe he forgot the RT =/= endorsement somewhere?

1120 12