Thomas Bloom

@thomasfbloom

Royal Society University Research Fellow at the University of Manchester. Mathematician and owner of https://t.co/SWVqqnq9hn. He/him/his.

5 382Followers

96Following

1 297Posts total

297.8KViews on collected posts

Against accounts of the same size

8 posts from the last 90 days, next to the under 10K follower range. shown to more people than peers of the same size.

Median views14 904this account2 112median for under 10K

Reach, %276.91%this account157.00%median for under 10K

Engagement, %1.24%this account1.62%median for under 10K

Metric	This account	Median for under 10K	Ratio
Median views per post	14 904	2 112	7.06×
Reach (views ÷ followers)	2.8× audience	157.00%	1.76×
Engagement rate	1.24%	1.62%	0.76×

Others in this range → Compare with another account → How these benchmarks are built →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

201.8K5 Sep

18.1K

16.2K

17.8K

12K

13.6K

10.9K

7.3K

Last 8 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.36%5 Sep

1.25%

1.07%

1.42%

1.51%

1.22%

1.64%

0.60%

Reactions — likes, reposts, replies and quotes — divided by views. Median for under 10K accounts is 1.62%.

What the audience does

Likes83.8%1 813 in total

Reposts2.4%52 in total

Replies2.8%61 in total

Quotes1.0%21 in total

Bookmarks10.0%217 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Latest posts

@thomasfbloom Widely known? Maybe to people with contacts at the AI companies, not by the actual mathematical community. It's encouraging that people that used to be very AI-friendly like Thomas or Terry Tao are finally realizing the huge dangers involved 7.3K views · 40 likes · 0 reposts · 4 replies 05 Sep 2026 · Open on X →

Is this a good policy? I'm genuinely unsure. Some days it feels like the best of a bunch of bad options, other days like the worst. Interested to hear opinions of others! 7/ 10.9K views · 168 likes · 1 reposts · 11 replies 05 Sep 2026 · Open on X →

Yet alternatives are also problematic - e.g. if they release floods of unverified AI-generated PDFs proving big conjectures. These will be papers that likely never go to peer review or publication, and can be field-killers, if there's no incentive for people to continue work. 5/ 13.6K views · 159 likes · 2 reposts · 4 replies 05 Sep 2026 · Open on X →

The current policy of OpenAI appears to be 'we've demonstrated that our AI can prove interesting stuff; now we'll let individuals explore for themselves, so the results come out at a slower pace, to not overwhelm, even though we've already done the proofs ourselves'. 6/ 12K views · 173 likes · 3 reposts · 4 replies 05 Sep 2026 · Open on X →

This situation is obviously discouraging for working mathematicians. Should one spend a long time thinking about a conjecture, if one suspects that there is a full formalised proof gathering dust in the servers of OpenAI/Anthropic? 4/ 17.8K views · 242 likes · 6 reposts · 5 replies 05 Sep 2026 · Open on X →

This is especially true of results with less mass appeal - there may be a few mathematicians who have devoted years to a conjecture, and care a lot about a solution, but otherwise the news has little impact, and might be mainly negative to those mathematicians. 3/ 16.2K views · 171 likes · 1 reposts · 1 replies 05 Sep 2026 · Open on X →

Some of these (e.g. bounded gaps between primes) are eventually released when the timing is convenient from a PR perspective. But now that the AI companies have demonstrated they can do interesting maths, their incentives drop off. 2/ 18.1K views · 224 likes · 1 reposts · 1 replies 05 Sep 2026 · Open on X →

It's now widely known that, along with the many big results announced, both OpenAI and Anthropic have many other results proved by their internal models that they have not released. Is this a good or bad thing? 1/ 201.8K views · 636 likes · 38 reposts · 31 replies 05 Sep 2026 · Open on X →