Telegram RegisterThe public register of Telegram
Telegram profile photo for Sparse Hash AI

Channel

Sparse Hash AI

@SparseHashAI

On this record: Topic · Growth · Engagement · Reactions · Posts · Citations · Cite this entry

175subscribers

+13 since we began measuring on 7 August 2026

Risers and fallers across the register · movement among entries of Under 1,000.

Register entry

Telegram ID-1001484837454
TypeChannel
Username@SparseHashAI
CreatedBetween 1 April 2019 and 30 September 2021 — estimated from Telegram’s id allocation, not measured. How this range is calculated.
First recorded8 August 2026
Last confirmed live7 October 2026
Measurements held7
Confirmed unchanged1 time, most recently 7 October 2026
On Telegramt.me/SparseHashAI

Topic

Technology — a classification, not a measurement. An on-box language model (Qwen3.6-35B-A3B-FP8, prompt version 1) read this channel’s own recent posts on 25 September 2026 and assigned it the closest of 31 fixed categories, at 99% confidence. This is a model’s judgement about what the channel is likely to be about, not a fact this register measured the way a subscriber count or a view count is measured — it can be revised on a later pass, and it carries no weight anywhere else on this page. How this classification works, and why it has no browse page of its own yet.

Growth

162175168.57 August 2026 — 162 subscribers8 August 2026 — 162 subscribers15 August 2026 — 163 subscribers30 August 2026 — 164 subscribers7 September 2026 — 163 subscribers26 September 2026 — 168 subscribers7 October 2026 — 175 subscribers7 August 20267 October 2026
7 measurements spanning 60 days, net +13. Dots are measurements; the straight line between them is drawn to join them, not to claim we know the path taken in between — snapshots are recorded only when a count changes, so gaps mean “no change observed”, never “interpolated”. The vertical axis spans 160–177 and does not start at zero.
Measurement log — every subscribers count we have recorded
Measured (UTC)SubscribersChange
7 Oct 2026, 02:59175+7
26 Sept 2026, 21:59168+5
7 Sept 2026, 08:19163-1
30 Aug 2026, 14:59164+1
15 Aug 2026, 19:53163+1
8 Aug 2026, 16:47162no change
7 Aug 2026, 17:12162first reading

Engagement

20 posts held, back to 14 July 2026 — the reader has not yet reached the start of this channel’s public history, so older posts may sit further back, unread. Read across 1 page of Telegram’s post history, 20 posts per page.

Nothing published in the last 30 days. ERR and ER are rolling 30-day measures, so there is nothing to compute — we hold 20 posts for this entry, the most recent from 6 August 2026. An engagement rate over an empty window would be a number about nothing.

Reaction mix

11 reactions across 9 posts, in 3 distinct kinds. The most used accounts for 81.8% of them.

Every reaction kind recorded on the sample, most used first
ReactionCountShareShare, drawn
🔥981.8%
❤19.09%
👍19.09%

No sentiment is inferred, and none should be read in. This table is ordered by count and by nothing else. Emoji do not carry stable meaning across languages or communities — 🙏 is thanks in one channel and mourning in another — so we publish which ones were pressed and how often, and pass no judgement on what an audience meant by them.

Precision. Telegram publishes reaction counts per emoji and short-forms each one — 4.34K, 1.2M — so any single kind at or above 1,000 reaches us at three significant figures, and only counts below 1,000 are exact. The shares above are ratios of those figures and carry the same error. This is also why the total here can differ slightly from a reaction total printed elsewhere on the page: both are sums of the same rounded parts, taken over samples with different edges.

Coverage. Reactions were read on 9 of the 20 sampled posts in this sample. Summed by Telegram’s own count on each post — not by adding up the per-emoji breakdown above — those same posts carry 11 reactions in total: the kind of figure the paragraph above means by “a reaction total printed elsewhere on the page”.

Measured over the 20 most recent posts we hold, published 14 July 2026 to 6 August 2026, using the newest reading held for each. Telegram Stars are excluded: they are a payment, not a reaction, and they have their own section.

Recent posts

6 Aug 2026, 08:53 UTC32 views1 reactionsread 8 August 2026
Photo

Proof of Concept Проверил также вариант, где значения не вычисляются из ключей, то есть имеем три классические отдельные проекции входа в ключи, запросы и значения. Схема другая, работает как ассоциативная память ключ-значение: 1. Связываем пары ключ-значение и собираем их суперпозицию – операция записи в память: ∑ⱼkⱼ⊙vⱼ. Используем cumsum для причинности. 2. Умножаем суперпозицию на запрос qᵢ – операция чтения …

❤1

6 Aug 2026, 08:10 UTC50 views3 reactionsread 8 August 2026
Photo

Proof of Concept Сделал проверку концепции суперпозиции концепций. Заменил в небольшом трансформере в self-attention вычисление оценок внимания и softmax на суперпозицию (ака векторная память). Для этого: 1. Собираем суперпозицию ключей: ∑ⱼkⱼ. Используем cumsum для причинности. 2, Умножаем суперпозицию на запрос qᵢ (произведение суперпозиций), это даёт суперпозицию всех связанных пар ключ-запрос: qᵢ⊙∑ⱼkⱼ 3. Про…

🔥3

4 Aug 2026, 12:16 UTC38 views1 reactionsread 8 August 2026

Матричная ассоциативная память (линейное внимание) и векторная ассоциативная память В линейном внимании в матрицу памяти S пара ключ-значение добавляется как внешнее произведение их векторов. S = S + kᵀv В векторную память пара ключ-значение добавляется как поэлементное произведение. S = S + k⊙v Чтение из линейного внимания – это матричное умножение ключа и состояния. o = kS Чтение из векторной памяти – это по…

🔥1

3 Aug 2026, 17:29 UTC62 viewsread 8 August 2026

Суперпозиция также представляет из себя ассоциативную память, из которой можно читать по запросу. Если суперпозиция содержит в себе пары концепций: [ac, ad, bc, bd], то умножив её на вектор запроса q = a, получим суперпозицию значений [c, d], где запрос является частью пары. (ac + ad + bc + bd) * a = c + d import torch import torch.nn.functional as F D = 1024 # Векторы концепций a, b, c, d = F.normalize(torch.ra…

3 Aug 2026, 16:57 UTC61 viewsread 8 August 2026

Если скрытое состояние представляет из себя суперпозицию, то возведение её в квадрат, например в ReLU², даст суперпозицию попарных взаимодействий элементов. S² = (a + b + c + d) * (a + b + c + d) = aa + 2ab + 2ac + 2ad + bb + 2bc + 2bd + cc + 2cd + dd import torch import torch.nn.functional as F D = 1024 # Векторы концепций a, b, c, d = F.normalize(torch.randn(4, D)) # Исходная суперпозиция S = a + b + c + d # …

3 Aug 2026, 16:10 UTC68 viewsread 8 August 2026

import torch import torch.nn.functional as F D = 1024 # Векторы концепций a, b, c, d = F.normalize(torch.randn(4, D)) # Исходные суперпозиции S1 = a + b S2 = c + d # Произведение суперпозиций S = S1 * S2 S = F.normalize(S.unsqueeze(0)).squeeze(0) # Пары концепций ac = a * c ad = a * d bc = b * c bd = b * d z = torch.vstack([ac, ad, bc, bd]) z = F.normalize(z) # Находим пары концепций в результирующей суперпози…

3 Aug 2026, 16:09 UTC63 views1 reactionsread 8 August 2026

💡 Произведение двух суперпозиций даёт суперпозицию попарных взаимодействий их элементов. Пример Пусть первая суперпозиция S1 содержит в себе концепции a и b, а вторая S2 - c и d. Перемножаем суперпозиции: S = S1 * S2 = (a + b) * (c + d) = ac + ad + bc + bd Результирующая суперпозиция S содержит в себе коды всех попарных комбинаций концепций. То есть получаем все возможные комбинации всего за одну операцию поэле…

👍1

3 Aug 2026, 13:37 UTC43 views1 reactionsread 8 August 2026

Weight-norm Criticality: A Mechanism for Loss Spikes Induced by the Normalization and Weight Decay Критичность нормы весов: Механизм скачков функции потерь, вызванных нормализацией и L2-регуляризацией https://www.alphaxiv.org/overview/2607.21005 Исследователи представляют "критичность нормы весов" — механизм, объясняющий всплески потерь в глубоких нейронных сетях, где взаимодействие слоев нормализации и уменьшения в…

🔥1

1 Aug 2026, 04:59 UTC61 views1 reactionsread 8 August 2026

Новый китайский стартап в области робототехники Light Origin. https://d.fixupx.com/ErenChenAI/status/2080638854311059759

🔥1

31 Jul 2026, 09:15 UTC68 viewsread 8 August 2026

When Does Recurrence Become an Algorithm? Convergence Selection in Weight-Tied Looped Transformers Когда рекуррентность становится алгоритмом? Выбор сходимости в зацикленных трансформерах со связанными весами https://www.alphaxiv.org/overview/2607.20594 Это исследование показывает, что трансформеры с циклическим выполнением и связанными весами реализуют «линейный вычислительный фронт» для решения алгоритмических зад…

30 Jul 2026, 23:03 UTC52 viewsread 8 August 2026

Полуавтономный робот для услуг по уборке в домах (робот на час) в Сан-Франциско. https://www.tau-robotics.com/ https://d.fixupx.com/CyberRobooo/status/2082271944976826720

Showing the 12 most recent of 20 posts we hold for @SparseHashAI. View and reaction counts are the latest single reading for each post, not a live figure, and a recent post is still accumulating both. A view count marked ≈ was rounded by Telegram before we ever saw it — t.me prints views in full below 1,000 and to three significant figures above, so ≈1,200,000 means somewhere between 1,150,000 and 1,249,999. Unmarked counts are exact. Text is reproduced from the public post preview and truncated for length.

Mentions

Names

Channels on the register whose handles appear in this channel's posts.

A mention is a weaker signal than a forward and is counted separately for that reason — naming a channel is not republishing it, and a handle in a post body is easy to place deliberately. The post counts beside each row below are distinct posts in which the handle appeared, from posts we have read on both sides — the “Named by N registered channels” figure above is a different count, of distinct NAMING CHANNELS rather than posts, and is not the sum of the rows under it.

Cite this entry

A live page changes as we take new readings, so a citation should name the measurement it is based on, not just the URL. The line below cites the subscriber count as measured 7 October 2026 — this entry's latest reading, not the date you are reading this.

“Sparse Hash AI” (@SparseHashAI), 175 subscribers as measured 7 October 2026. Telegram Register, tgregister.com/channel/SparseHashAI.

Full measurement history, CC BY 4.0. Every reading this register holds for this entry, not just the latest one, as a dated, downloadable record: CSV · JSON. Free to use with attribution to tgregister.com. Each file carries its own generation timestamp, which is the figure to cite for exactly when the data was retrieved.