DEV Community: John

AI signals this week (2026-10-26 to 2026-11-01): what actually moved

John — Sun, 01 Nov 2026 15:00:04 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2390 items collected across 7 days and 7 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 312 of 2149 tracked, as of 2026-11-01)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 312 by Hugging Face downloads among the entries we track; downloads -4.1% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 312 by Hugging Face downloads among the entries we track; downloads -4.7% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 312 by Hugging Face downloads among the entries we track; downloads -5.8% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 76 of 76 tracked, as of 2026-11-01)

codecrafters-io/build-your-own-x: ranked #1 of 76 by GitHub stars among the entries we track; stars +0.3% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 76 by GitHub stars among the entries we track; stars +0.7% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 76 by GitHub stars among the entries we track; stars +0.4% since 2026-10-25 (as of 2026-11-01). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 4 new entries, 37 not seen in the latest snapshot, 306 rank moves

mistralai/Mistral-Large-4-1T-A52B moved from #124 to #116 (+55.4% downloads)
baidu/Unlimited-OCR-2 moved from #66 to #60 (+58.2% downloads)
ccore/Qwen3.5-4B-3D-Coder moved from #78 to #73 (+51.4% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`openai-community/gpt2`	+4,151,730	24,837,303 → 28,989,033
`Comfy-Org/MiniMax-H3`	+3,707,117	26,539,748 → 30,246,865
`ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF`	+1,185,477	891,147 → 2,076,624
`hexgrad/Kokoro-82M`	+912,177	10,335,335 → 11,247,512
`BAAI/bge-m3`	+567,242	29,783,764 → 30,351,006

Biggest repo gains (stars)

Repo	Gain	From → To
`morluto/rea`	+32,463	384,462 → 416,925
`Gathin23/cleanyourtv`	+13,486	34,903 → 48,389
`sindresorhus/awesome`	+3,584	524,651 → 528,235
`wheresryan22/anatomy`	+2,978	20,577 → 23,555
`deepseek-ai/DeepInfer`	+2,363	22,380 → 24,743

What showed up, by source

Silicon Valley news (189 items)

Industry news (1005 items)

AI hardware (175 items)

AMD and Meta Advance Strategic Partnership to Power Meta’s Superintelligence Infrastructure — 2026-11-01
Introducing AMD Helion, Agentic AI Compute at Rack Scale — 2026-11-01
AMD to Report Fiscal Third Quarter 2026 Financial Results — 2026-11-01
AMD to Acquire World Labs to Advance the Future of AI Compute — 2026-11-01

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

AI signals this week (2026-10-19 to 2026-10-25): what actually moved

John — Sun, 25 Oct 2026 15:00:05 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2423 items collected across 7 days and 7 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 281 of 1906 tracked, as of 2026-10-25)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 281 by Hugging Face downloads among the entries we track; downloads -8.3% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 281 by Hugging Face downloads among the entries we track; downloads -7.4% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 281 by Hugging Face downloads among the entries we track; downloads -7.7% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 70 of 70 tracked, as of 2026-10-25)

codecrafters-io/build-your-own-x: ranked #1 of 70 by GitHub stars among the entries we track; stars +0.3% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 70 by GitHub stars among the entries we track; stars +0.7% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 70 by GitHub stars among the entries we track; stars +0.4% since 2026-10-18 (as of 2026-10-25). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 1 new entries, 39 not seen in the latest snapshot, 186 rank moves

XHToken/Spark-X2.6-4B moved from #115 to #108 (+37.5% downloads)
MiniMaxAI/MiniMax-M3.1-Flash moved from #104 to #100 (+27.1% downloads)
deepseek-ai/DeepSeek-V4.1-Flash-1017 moved from #64 to #61 (+16.5% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`Comfy-Org/MiniMax-H3`	+1,297,452	25,242,296 → 26,539,748
`google-t5/t5-small`	+550,085	24,230,244 → 24,780,329
`black-forest-labs/flux-3-image-klein-comfy`	+322,838	338,036 → 660,874
`deepseek-ai/DeepSeek-R1`	+175,409	1,310,380 → 1,485,789
`autotrust/JEV-27B-VL`	+128,476	1,801,077 → 1,929,553

Biggest repo gains (stars)

Repo	Gain	From → To
`morluto/rea`	+44,976	336,331 → 381,307
`Ebony-Vinyl/dsh-our-free-model`	+14,373	24,711 → 39,084
`Gathin23/cleanyourtv`	+14,265	18,944 → 33,209
`storytold/photocraft`	+5,158	67,433 → 72,591
`wheresryan22/anatomy`	+4,399	15,979 → 20,378

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-10-25
sindresorhus/awesome — 2026-10-25
public-apis/public-apis — 2026-10-25
freeCodeCamp/freeCodeCamp — 2026-10-25

Hugging Face (341 items)

deepseek-ai/DeepSeek-V4.1-Flash-1017 — 2026-10-25
RecursiveSI/flint — 2026-10-25
black-forest-labs/flux-3-image-klein-comfy — 2026-10-25
XHToken/Spark-X2.6-4B — 2026-10-25

arXiv (446 items)

Direct Pass@$k$ Optimization for Thought Pruning: Balancing Quality and Diversity in Test-Time Reasoning — 2026-10-25
Cost-Aware Abstention for LLM-Based IT Ticket Classification — 2026-10-25
METTA: Memory-Enhanced Test-Time Adaptation for Meta-Reinforcement Learning — 2026-10-25
EpiSEM-Bench: Evaluating Safety and Evidence Recovery in LLM Agents under Missing Evidence — 2026-10-25

YouTube (84 items)

Do Chip Fabs Even Need Clean Rooms? — 2026-10-25
There Aren’t Enough Machines to 10x AI Chips — 2026-10-25
Gemini 4 Argon Is Not What I Expected — 2026-10-25
Why Russia’s Tsar Lost Control - Sarah Paine — 2026-10-25

Silicon Valley news (193 items)

Industry news (1044 items)

AI hardware (175 items)

Introducing AMD Helion, Agentic AI Compute at Rack Scale — 2026-10-25
AMD to Report Fiscal Third Quarter 2026 Financial Results — 2026-10-25
AMD to Acquire World Labs to Advance the Future of AI Compute — 2026-10-25
AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-10-25

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

AI signals this week (2026-10-12 to 2026-10-18): what actually moved

John — Sun, 18 Oct 2026 15:00:04 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2579 items collected across 7 days and 8 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 255 of 1669 tracked, as of 2026-10-18)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 255 by Hugging Face downloads among the entries we track; downloads -6.4% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 255 by Hugging Face downloads among the entries we track; downloads -3.7% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 255 by Hugging Face downloads among the entries we track; downloads -3.4% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 66 of 66 tracked, as of 2026-10-18)

codecrafters-io/build-your-own-x: ranked #1 of 66 by GitHub stars among the entries we track; stars +0.3% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 66 by GitHub stars among the entries we track; stars +0.6% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 66 by GitHub stars among the entries we track; stars +0.5% since 2026-10-11 (as of 2026-10-18). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 5 new entries, 42 not seen in the latest snapshot, 214 rank moves

deepseek-ai/DeepSeek-V4.1-Flash-1017 moved from #127 to #94 (+1089.9% downloads)
openbmb/MiniCPM-V-4.7-35B-A3B moved from #69 to #63 (+59.8% downloads)
bropal/lma-project-nepali moved from #231 to #236 (+0.0% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`Comfy-Org/MiniMax-H3`	+1,661,404	23,580,892 → 25,242,296
`Qwen/Qwen3.8-27B`	+952,435	6,843,108 → 7,795,543
`abenzerps/Qwen-Image-2.1-Uncensored-Quantized`	+383,064	2,188,658 → 2,571,722
`Lightricks/LTX-2.5`	+285,813	1,726,683 → 2,012,496
`deepseek-ai/DeepSeek-R1`	+190,619	1,119,761 → 1,310,380

Biggest repo gains (stars)

Repo	Gain	From → To
`storytold/photocraft`	+19,460	47,434 → 66,894
`Ebony-Vinyl/dsh-our-free-model`	+11,719	11,870 → 23,589
`storytold/pdfcraft`	+6,094	9,581 → 15,675
`Niko1221/Strata`	+5,586	21,564 → 27,150
`storytold/filmcraft`	+4,637	10,054 → 14,691

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-10-18
sindresorhus/awesome — 2026-10-18
public-apis/public-apis — 2026-10-18
freeCodeCamp/freeCodeCamp — 2026-10-18

Hugging Face (341 items)

deepseek-ai/DeepSeek-V4.1-Flash-1017 — 2026-10-18
google/embeddinggemma-2 — 2026-10-18
jialinyyzz/humanizer — 2026-10-18
openbmb/MiniCPM-V-4.7-35B-A3B — 2026-10-18

arXiv (423 items)

LIBERO-PeRM: Benchmarking Personalized Robotic Manipulation — 2026-10-18
Discovering state-of-the-art quantum circuits with large language models — 2026-10-18
T2-z: Test-Time Optimization of Environment Latents for Robotic World Models — 2026-10-18
CoDA: Co-Evolution of Data Engines and Action Policies — 2026-10-18

LMSYS Arena (141 items)

gemini-4-argon-high — 2026-10-14
claude-opus-4-6-high — 2026-10-14
claude-fable-5-high — 2026-10-14
claude-opus-5.5-high — 2026-10-14

YouTube (117 items)

Jensen Huang on China's AI labs | Lex Fridman in Taiwan — 2026-10-18
Jensen Huang reflects on huge crowds greeting him in Taiwan | Lex Fridman — 2026-10-18
The best street food in Taiwan with Jensen Huang | Lex Fridman — 2026-10-18
Jensen Huang on China, GPUs, and export controls | Lex Fridman in Taiwan — 2026-10-18

Silicon Valley news (193 items)

An interview with Apple's new VP of the App Store, Carson Oliver, on Apple's gaming vision, Apple Arcade, providing tools for indie game developers, and more (Giovanni Colantonio/Polygon) — 2026-10-18
Kids are outsmarting social media bans. That's not a problem we need to solve — 2026-10-18
How employment is changing at Anthropic — 2026-10-18
It's time to talk about TEEs — 2026-10-18

Industry news (1049 items)

AI hardware (175 items)

Introducing AMD Helion, Agentic AI Compute at Rack Scale — 2026-10-18
AMD to Report Fiscal Third Quarter 2026 Financial Results — 2026-10-18
AMD to Acquire World Labs to Advance the Future of AI Compute — 2026-10-18
AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-10-18

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

44 days of AI signals as a dataset — and 3 silent API traps I hit publishing it

John — Wed, 14 Oct 2026 21:14:03 +0000

I've been collecting AI signals from public sources every day for 44 days. This week I published
the whole thing as a machine-readable dataset. Along the way I hit three failure modes that were
completely silent — no error, no log line, just "nothing happens". They're worth knowing if you
publish a public API or dataset, because they cost me hours and I'd rather you skip them.

What the dataset is

17,397 rows, one row per signal, across 44 daily editions (2026-09-01 → 2026-10-14)
Fields: source, board, title, url, summary, org, topics, timestamp, edition_date
Sources: GitHub trending, Hugging Face models, YouTube talks, arXiv papers, LMSYS Arena boards, AI news
JSONL, CC BY 4.0 (commercial use fine, just credit), refreshed daily
https://huggingface.co/datasets/leopardyzzj/ai-signals-daily

Trap 1 — Your CDN 403s stdlib clients, and your logs show nothing

If your JSON endpoints sit behind Cloudflare, a request with the default Python-urllib/3.x
User-Agent (or an empty UA) can get a 403 at the edge. The request never reaches your origin,
so your access log is empty and your monitoring is green.

The trap isn't the 403 — it's the conclusion you draw from it. You look at your logs, see no traffic,
and decide "nobody is using my API." Meanwhile people following the standard-library examples in
your docs are getting 403s and assuming your API is broken.

Fix: send a real User-Agent. Then document it, because the default behaviour of urllib is exactly
the thing being blocked:

req = urllib.request.Request(url)
req.add_header("User-Agent", "my-app/1.0 (+https://example.com)")

requests, httpx, aiohttp, Go's client, node-fetch, axios, okhttp and curl all pass
fine — it's specifically the bare stdlib default that gets filtered.

Trap 2 — Hugging Face's per-file upload endpoint is gone (410)

I built an uploader against POST /api/datasets/{repo}/upload/main/{filename}. It used to work.
Now it returns:

HTTP 410: This endpoint is deprecated and has been retired, use the commit endpoint instead.

The replacement is the commit endpoint, which takes NDJSON: a header line, then one line per
file with base64 content. Small files go inline; anything past the 10MB mark needs LFS.

lines = [json.dumps({"key": "header", "value": {"summary": "daily refresh"}})]
for f in files:
    lines.append(json.dumps({"key": "file", "value": {
        "path": f.name,
        "encoding": "base64",
        "content": base64.b64encode(f.read_bytes()).decode(),
    }}))
req = urllib.request.Request(
    f"https://huggingface.co/api/datasets/{repo}/commit/main",
    data=("\n".join(lines) + "\n").encode(),
    method="POST",
    headers={"Authorization": f"Bearer {token}",
             "Content-Type": "application/x-ndjson"})

Trap 3 — Don't hardcode the namespace

This one is embarrassing and it's the one that would have wasted the most time.

My publisher had owner = "my-org-name" written into it, but the API token I was handed belonged to a
different account. The result: the repository got created under the token's account, and every
file upload was sent to the other namespace. Every single file returned 403, and it looks exactly
like a flaky network problem.

Derive the namespace from the token instead, and verify the token actually has write access before
you start a 200-file upload:

# GET https://huggingface.co/api/whoami-v2
# -> {"name": "<account>", "auth": {"accessToken": {"role": "write"}}}

If that role isn't write, a read-only token will get you authenticated, look fine, and then fail
at upload time — which is the worst place to discover it.

The common thread

All three traps share one property: they fail silently, and the misleading signal arrives in a
different place than the failure. A CDN block shows up as "no traffic". A dead endpoint shows up as
"403 from the namespace". A read-only token shows up as an authentication error.

If you're publishing data or an API, here's the cheap insurance: probe your own surface anonymously,
from outside, with the same defaults a stranger would use — and do it after every deploy. Most of my
worst bugs were invisible from the inside.

The dataset: huggingface.co/datasets/leopardyzzj/ai-signals-daily
(CC BY 4.0). There's also a free daily digest (4 signals, one per core source), and if you want the
whole archive as JSONL in one go, it's a one-time CA$5:
buy.stripe.com/00w28r4W08qq8CcecHgw008.

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

43 days, 17,040 AI signals: four things the data says that trend lists won't tell you

John — Wed, 14 Oct 2026 04:00:07 +0000

I've been running a small pipeline that collects AI signals from public sources for 43 days now — 17,040 items, ~398 a day. I built it to power a daily digest, but the dataset turned out to be more interesting than the digest, mostly because of what it can't tell me.

Here are the four findings I'd want to know before building anything on top of AI trend data.

1. Only 23 of 1,493 tracked models have enough history to compute a trend

This is the big one. My collector snapshots trending lists — GitHub trending, Hugging Face trending, and so on. Those lists rotate. A model that's #4 today may be gone tomorrow, taking its history with it.

The result:

	tracked	ranked	with ≥20 days of history
Hugging Face models	1,493	228	23
GitHub repos	63	63	14

So when a tool shows you "momentum" for an AI model, ask what its baseline is. If the entity was only observed twice, a "trend" is two points and a straight line.

I chose to publish unranked with the reason rather than fill the gap with something plausible — every row in my JSON either has a derivable number or an explicit "not ranked, here's why". That decision cost me a lot of "wow" numbers. It's still the right one: an invented rank is indistinguishable from a real one to whatever consumes it downstream.

The fix, if you're building this yourself: track a fixed set (top N by a stable metric, every day) in addition to the rotating trending lists. Rotation gives you discovery; a fixed set gives you a time series. Without the second one you have neither.

2. Top entities are boringly stable — the interesting signal is churn

I expected the leaders to move. Over 48 days of daily snapshots:

Biggest Hugging Face download change: −11.7% (moonshotai/Kimi-K3)
Biggest increase in the same column: +7.6% (deepseek-ai/DeepSeek-R1)
Everything else in the top 8: −5% to −10%
GitHub stars, top repos: +0.08% to +0.54% over 48 days

So the leaderboard is nearly static, and most large models slowly decay in downloads. Growth is the exception, not the norm.

Meanwhile, a single day's deltas (2026-10-12 → 2026-10-13) looked like this: 2 models entered, 41 were not seen again, 201 changed rank. That's where the motion is — at the edges, not the top.

If you're building alerts, alert on entry and exit, not on rank movement at the top. "This model appeared for the first time / this one vanished" carried far more information than "+3 places".

3. Most of the volume is one source, and it isn't the interesting one

Over 43 days, by source:

source	items	share
hot-topic aggregation	6,455	38%
arXiv	2,551	15%
Hugging Face	2,107	12%
arena boards	2,041	12%
silicon-valley news	1,198	7%
hardware/chips	1,089	6%
GitHub	860	5%
YouTube	739	4%

The aggregation source alone is more than a third of everything. It's cheap to collect and it mostly duplicates what you already have from the primary sources.

The four sources that actually produce the digest (GitHub, Hugging Face, YouTube, Silicon Valley news) are 29% of the volume. If you rank sources by item count, you'll optimize for the loudest one instead of the useful one. I ended up tracking per-source yield and dedup rate, not raw volume — volume is the metric that makes a pipeline look productive while producing nothing.

4. Derive numbers from raw snapshots, or don't publish them

Every derived number I publish (rank, rank change, 7-day momentum, position in an arXiv category) is recomputed from raw snapshots by an automated audit, and the audit also reconciles the rendered pages against the JSON endpoints. If the two disagree, that's a failure, not a rounding difference.

Two reasons this matters more than it sounds:

Derived numbers drift silently. A rank computed at write time and stored is a claim from the past; recomputed, it's a claim you can defend.
It catches your own bugs. The audit has caught things like a "content quality" check passing on pages whose body was actually empty, and mismatch between a daily edition and the API response for the same date.

I'd rather ship a boring number I can re-derive than an exciting one I can't.

What I'm doing with this

The pipeline is unattended (collects 4×/day, one published edition/day, self-audited daily). The data is public and free:

Daily dataset (CC BY 4.0): github.com/jiangzhangcc-glitch/ai-signals-daily — plus a bulk signals.jsonl (all 17,040 rows) on the releases page
Today's signals as JSON: https://data.jiangzhang.ca/signals4/api/today.json
What changed since the previous snapshot: .../api/changes.json
Ranked models/repos with citations: .../api/models.json, .../api/repos.json
As MCP tools for Claude/Cursor: signals-api-mcp

One gotcha worth passing on: if you call the API with a bare Python-urllib User-Agent (or none), you'll get a 403 from the CDN, at the edge — your request never reaches the server, and the server's logs show nothing. It looks exactly like an API that's down. requests, httpx, aiohttp, Go, node-fetch, axios and curl are all fine. Set a User-Agent in examples.

If you need more than the free tier — 8 sources instead of 4, four updates a day, and full history behind a stable key — that's the paid feed: what it includes.

And if you're building something similar: I'd genuinely like to know how you handle the rotating-list problem. It's the part I underestimated most.

Building an MCP server for AI signals data — and the Cloudflare 403 that silently broke it

John — Wed, 14 Oct 2026 03:21:11 +0000

I wanted my coding agent to answer "what's happening in AI today?" without me pasting links into the chat. So I wrote a small MCP server for the AI signals data I collect. Two things came out of it that might save you time.

1. The MCP server (single file, stdlib only, no API key)

Five tools, all read-only:

Tool	Returns
`daily_signals`	today's signals — one per source (GitHub, Hugging Face, YouTube, Silicon Valley news), each linking to the original
`what_changed`	deltas since the previous snapshot: new entries, entities no longer seen, rank movers
`top_models`	Hugging Face models with rank, rank movement, momentum
`top_repos`	GitHub repos, same shape
`archive_days`	which daily editions exist, so the agent can cite a date

Every ranked row includes a citation string, so the agent cites a number instead of inventing one. Rows whose number can't be derived come back as unranked with the reason — I'd rather return "not ranked" than a plausible-looking guess.

Install is one file plus a config block:

{
  "mcpServers": {
    "signals": {
      "command": "python3",
      "args": ["/absolute/path/to/mcp_server.py"]
    }
  }
}

Repo: https://github.com/jiangzhangcc-glitch/signals-api-mcp (MIT, Python 3.9+, no dependencies)

2. The trap: a 403 that your server never sees

My first version called the API with plain urllib.request.urlopen(url). Every tool failed with HTTP 403. The API was fine — the CDN in front of it rejects the bare Python-urllib/… User-Agent (and empty User-Agents) before the request ever reaches the origin.

The nasty part: because the request is blocked at the edge, the origin server has no log of it. If you only look at your own access logs, you conclude "nobody is calling my API". I measured it across clients:

User-Agent	Result
`Python-urllib/3.12` (stdlib default)	403
(empty)	403
`python-requests/2.31.0`	200
`aiohttp`, `httpx`, `Go-http-client`, `node-fetch`, `axios`, `okhttp`, `Java`	200
`curl/8.5.0`	200

So requests is fine — but the zero-dependency example that a reader will copy-paste is exactly the one that breaks. If you publish an API, test it with the dumbest client you can write, and set a real User-Agent in every sample:

curl -H 'User-Agent: my-app/1.0' https://data.jiangzhang.ca/signals4/api/today.json

The server now sends SignalsAPI-MCP/1.0 (+https://…) on every request, and the 403 case returns an error message that names the likely cause instead of a bare status code.

The data behind it

I run a small daily pipeline that collects from public sources (GitHub trending, Hugging Face, YouTube, arXiv, arena boards, tech news, hardware/chips), dedupes and archives it. The newest daily delta is public, plus a machine-readable dataset:

today's signals: https://data.jiangzhang.ca/signals4/api/today.json
daily dataset mirror: https://github.com/jiangzhangcc-glitch/ai-signals-daily (CC BY 4.0)
site summary for agents: https://data.jiangzhang.ca/llms.txt

The free tier is one edition a day from four core sources. There's a paid feed if you need eight sources, four updates a day, and the full history as an API — it's linked from the repo README. Everything in the MCP server stays free.

If you build something with it, I'd genuinely like to know what breaks — the 403 above only surfaced because I used my own server as a user.

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

AI signals this week (2026-10-05 to 2026-10-11): what actually moved

John — Sun, 11 Oct 2026 15:00:04 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2734 items collected across 7 days and 8 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 219 of 1426 tracked, as of 2026-10-11)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 219 by Hugging Face downloads among the entries we track; downloads -5.9% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 219 by Hugging Face downloads among the entries we track; downloads -5.6% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 219 by Hugging Face downloads among the entries we track; downloads -4.0% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 61 of 61 tracked, as of 2026-10-11)

codecrafters-io/build-your-own-x: ranked #1 of 61 by GitHub stars among the entries we track; stars +0.2% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 61 by GitHub stars among the entries we track; stars +0.5% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 61 by GitHub stars among the entries we track; stars +0.3% since 2026-10-04 (as of 2026-10-11). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 2 new entries, 37 not seen in the latest snapshot, 183 rank moves

google/embeddinggemma-2 moved from #58 to #50 (+116.7% downloads)
Qwen/Qwen-Image-2.1-Turbo moved from #93 to #87 (+119.9% downloads)
Aleph-Alpha/Kolibri-1 moved from #80 to #77 (+19.6% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`Qwen/Qwen3-0.6B`	+1,102,866	29,665,992 → 30,768,858
`abenzerps/Qwen-Image-2.1-Uncensored-GGUF`	+634,914	1,553,744 → 2,188,658
`autotrust/GEV-26B-Decide`	+463,228	446,527 → 909,755
`Comfy-Org/MiniMax-H3`	+324,069	23,256,823 → 23,580,892
`autotrust/JEV-27B-VL`	+257,964	1,278,569 → 1,536,533

Biggest repo gains (stars)

Repo	Gain	From → To
`storytold/photocraft`	+27,746	16,749 → 44,495
`Niko1221/Strata`	+10,204	10,902 → 21,106
`robbietilton/Compositor`	+8,150	7,558 → 15,708
`eternity4719/HowToLiveBetter`	+6,216	40,361 → 46,577
`openai/math`	+4,519	9,346 → 13,865

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-10-11
sindresorhus/awesome — 2026-10-11
public-apis/public-apis — 2026-10-11
freeCodeCamp/freeCodeCamp — 2026-10-11

Hugging Face (340 items)

google/embeddinggemma-2 — 2026-10-11
jialinyyzz/humanizer — 2026-10-11
Cloudflare/clef — 2026-10-11
abenzerps/Qwen-Image-2.1-Uncensored-GGUF — 2026-10-11

arXiv (439 items)

On the estimation and validity of AI time horizons---a statistical look at the METR plot — 2026-10-11
From Reactive Containment to Proactive Assurance: Lessons from OpenAI, Anthropic, and Google Agent Security Incidents — 2026-10-11
BrickBench: Evaluating Agentic Brick Design — 2026-10-11
Bi-FORK: Generative Modeling of High-Dimensional Bifurcating Systems — 2026-10-11

LMSYS Arena (327 items)

gemini-4-argon-high — 2026-10-11
claude-opus-4-6-high — 2026-10-11
claude-fable-5-high — 2026-10-11
claude-opus-5.5-high — 2026-10-11

YouTube (69 items)

Why the conquistadors thought they could win — Si Sheppard — 2026-10-11
Why the Aztecs were doomed - Si Sheppard — 2026-10-11
Why the Maya were harder to conquer than the Aztecs — Si Sheppard — 2026-10-11
What the Aztec and Inca emperors got wrong about the Spanish — Si Sheppard — 2026-10-11

Silicon Valley news (193 items)

Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things — 2026-10-11
Nitter: Update Oct 10th seeking funding and legal help — 2026-10-11
"Robot" Is a Social Construct: Why Your Dishwasher Is Not a Robot — 2026-10-11
IRCv3 — 2026-10-11

Industry news (1051 items)

Nvidia in talks to invest further in Reflection AI or buy it, FT reports - Reuters — 2026-10-11
Nvidia in talks to acquire US ‘open’ model start-up Reflection AI - Financial Times — 2026-10-11
Nvidia Discusses Boosting Reflection Investment, Computing Deal - Bloomberg.com — 2026-10-11
Super Micro contractor pleads guilty in scheme to divert AI servers with Nvidia chips to China - Reuters — 2026-10-11

AI hardware (175 items)

AMD to Report Fiscal Third Quarter 2026 Financial Results — 2026-10-11
AMD to Acquire World Labs to Advance the Future of AI Compute — 2026-10-11
AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-10-11
AMD Appoints Tim Ryan to Board of Directors — 2026-10-11

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

AI signals this week (2026-09-28 to 2026-10-04): what actually moved

John — Sun, 04 Oct 2026 15:00:03 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2690 items collected across 7 days and 8 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 181 of 1191 tracked, as of 2026-10-04)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 181 by Hugging Face downloads among the entries we track; downloads -2.8% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 181 by Hugging Face downloads among the entries we track; downloads -2.5% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 181 by Hugging Face downloads among the entries we track; downloads -0.9% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 56 of 56 tracked, as of 2026-10-04)

codecrafters-io/build-your-own-x: ranked #1 of 56 by GitHub stars among the entries we track; stars +0.2% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 56 by GitHub stars among the entries we track; stars +0.6% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 56 by GitHub stars among the entries we track; stars +0.4% since 2026-09-27 (as of 2026-10-04). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 3 new entries, 40 not seen in the latest snapshot, 133 rank moves

convaiinnovations/laya moved from #90 to #78 (+303.9% downloads)
Cloudflare/clef moved from #81 to #77 (+60.8% downloads)
ReliquaryForge/qwen3-4b-base-dapo-v4 moved from #40 to #44 (-53.5% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2`	+6,572,462	45,390,285 → 51,962,747
`Comfy-Org/MiniMax-H3`	+1,398,948	21,857,875 → 23,256,823
`amazon/chronos-2`	+879,784	21,985,629 → 22,865,413
`stabilityai/stable-diffusion-xl-base-1.0`	+627,025	3,658,757 → 4,285,782
`abenzerps/Qwen-Image-2.1-Uncensored-GGUF`	+589,524	964,220 → 1,553,744

Biggest repo gains (stars)

Repo	Gain	From → To
`eternity4719/HowToLiveBetter`	+18,263	20,593 → 38,856
`NandhaKishorM/laya`	+3,882	26,688 → 30,570
`sindresorhus/awesome`	+2,950	511,472 → 514,422
`public-apis/public-apis`	+2,165	483,819 → 485,984
`vinta/awesome-python`	+1,499	323,553 → 325,052

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-10-04
sindresorhus/awesome — 2026-10-04
public-apis/public-apis — 2026-10-04
freeCodeCamp/freeCodeCamp — 2026-10-04

Hugging Face (345 items)

Cloudflare/clef — 2026-10-04
convaiinnovations/laya — 2026-10-04
abenzerps/Qwen-Image-2.1-Uncensored-GGUF — 2026-10-04
Lightricks/LTX-2.5 — 2026-10-04

arXiv (409 items)

One Basis to Animate Them All: Gaussian Blendshape Distillation for Real-Time Avatars — 2026-10-04
KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards — 2026-10-04
Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents — 2026-10-04
ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research — 2026-10-04

LMSYS Arena (332 items)

gemini-4-argon-high — 2026-10-04
claude-opus-4-6-high — 2026-10-04
claude-fable-5-high — 2026-10-04
claude-opus-5.5-high — 2026-10-04

YouTube (52 items)

Why Spain’s New World Riches Backfired - Si Sheppard — 2026-10-04
How 168 Spaniards Captured the Inca Emperor - Si Sheppard — 2026-10-04
The Mistake Both Sides Made in the Korean War - Sarah Paine — 2026-10-04
Korea Shows Why Every War Needs a Clear Goal - Sarah Paine — 2026-10-04

Silicon Valley news (195 items)

Industry news (1042 items)

NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI - NVIDIA Blog — 2026-10-04
Nvidia's bet that its chips can finance the AI boom gets a Wall Street reality check - Reuters — 2026-10-04
3 AI Chip Stocks To Watch After Nvidia Stock Hit A Fresh All Time High - Yahoo Finance — 2026-10-04
The 7-year-old Nvidia Shield TV is now $100 more expensive due to AI - Ars Technica — 2026-10-04

AI hardware (175 items)

AMD to Acquire World Labs to Advance the Future of AI Compute — 2026-10-04
AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-10-04
AMD Appoints Tim Ryan to Board of Directors — 2026-10-04
AMD Acquires Taalas to Advance Compute Solutions for Rapidly Growing AI Inference Market — 2026-10-04

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

AI signals this week (2026-09-21 to 2026-09-27): what actually moved

John — Sun, 27 Sep 2026 15:00:03 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2819 items collected across 7 days and 8 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 151 of 959 tracked, as of 2026-09-27)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 151 by Hugging Face downloads among the entries we track; downloads -3.3% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 151 by Hugging Face downloads among the entries we track; downloads -2.2% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 151 by Hugging Face downloads among the entries we track; downloads -1.9% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 52 of 52 tracked, as of 2026-09-27)

codecrafters-io/build-your-own-x: ranked #1 of 52 by GitHub stars among the entries we track; stars +0.3% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 52 by GitHub stars among the entries we track; stars +0.6% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 52 by GitHub stars among the entries we track; stars +0.4% since 2026-09-20 (as of 2026-09-27). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 2 new entries, 38 not seen in the latest snapshot, 144 rank moves

Edge0/Audio8-ASR-Infinite moved from #65 to #60 (+147.3% downloads)
convaiinnovations/laya moved from #70 to #73 (-5.7% downloads)
XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B moved from #64 to #66 (+11.8% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`Qwen/Qwen3-0.6B`	+5,597,305	23,745,227 → 29,342,532
`Comfy-Org/Qwen-Image-2.1`	+2,557,448	1,429,925 → 3,987,373
`prism-ml/Ternary-Bonsai-2-27B-gguf`	+1,435,352	1,908,396 → 3,343,748
`abenzerps/Qwen-Image-2.1-Uncensored-GGUF`	+781,907	182,313 → 964,220
`stabilityai/stable-diffusion-xl-base-1.0`	+570,247	3,088,510 → 3,658,757

Biggest repo gains (stars)

Repo	Gain	From → To
`NandhaKishorM/laya`	+16,240	10,010 → 26,250
`eternity4719/HowToLiveBetter`	+10,965	8,507 → 19,472
`browser-use/jev-ultrafast`	+8,928	11,752 → 20,680
`sindresorhus/awesome`	+2,941	508,245 → 511,186
`tamaratran/fast-jev-compaction`	+1,864	5,127 → 6,991

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-09-27
sindresorhus/awesome — 2026-09-27
public-apis/public-apis — 2026-09-27
freeCodeCamp/freeCodeCamp — 2026-09-27

Hugging Face (346 items)

convaiinnovations/laya — 2026-09-27
Qwen/Qwen-Image-2.1 — 2026-09-27
abenzerps/Qwen-Image-2.1-Uncensored-GGUF — 2026-09-27
XingChen-AGI/Xing4.0-29B-A4B — 2026-09-27

arXiv (426 items)

LLM Agents Can Easily Tamper With Their Own Traces — 2026-09-27
AD-WM: Action-Discriminative World Models for Counterfactual Model Predictive Control — 2026-09-27
RAPID: Robot Agentic Programming from Demonstrations — 2026-09-27
Rolling-WAM: World Action Models with Rolling Imagination — 2026-09-27

LMSYS Arena (328 items)

claude-opus-5.5-high — 2026-09-27
claude-opus-4-6-high — 2026-09-27
claude-fable-5-high — 2026-09-27
claude-opus-4-7-high — 2026-09-27

YouTube (157 items)

How Japan's invasion of China led to Pearl Harbor - Sarah Paine — 2026-09-27
What the Hugging Face Incident Reveals About AI Alignment - Noam Brown — 2026-09-27
Claude Opus 5.5 AI: An Incredible Leap Forward — 2026-09-27
What If AI Progress Doesn’t Slow Down? - Noam Brown — 2026-09-27

Silicon Valley news (195 items)

Industry news (1052 items)

Jensen Huang talks about AI and climate change like a supervillain - The Verge — 2026-09-27
Thieves Stole ‘Nvidia’ Trailers. They Got 20 Tons of Sand - WIRED — 2026-09-27
AMD vs. Nvidia: Which AI Stock Is the Better Buy? - Yahoo Finance — 2026-09-27
Student protesters disrupt NVIDIA AI climate panel - Al Jazeera — 2026-09-27

AI hardware (175 items)

AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-09-27
AMD Appoints Tim Ryan to Board of Directors — 2026-09-27
AMD Acquires Taalas to Advance Compute Solutions for Rapidly Growing AI Inference Market — 2026-09-27
AMD Reports Second Quarter 2026 Financial Results — 2026-09-27

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

AI signals this week (2026-09-14 to 2026-09-20): what actually moved

John — Sun, 20 Sep 2026 15:00:04 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2781 items collected across 7 days and 8 sources.

The standing numbers (from our own endpoints)

These are rankings we compute from our own daily snapshots — the same numbers as the site and the JSON endpoints, free to quote with attribution.

Models (ranked 119 of 716 tracked, as of 2026-09-20)

sentence-transformers/all-MiniLM-L6-v2: ranked #1 of 119 by Hugging Face downloads among the entries we track; downloads -0.3% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/sentence-transformers-all-minilm-l6-v2.html
cross-encoder/ms-marco-MiniLM-L6-v2: ranked #2 of 119 by Hugging Face downloads among the entries we track; downloads +1.3% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/cross-encoder-ms-marco-minilm-l6-v2.html
BAAI/bge-small-en-v1.5: ranked #3 of 119 by Hugging Face downloads among the entries we track; downloads +0.7% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/models/baai-bge-small-en-v1-5.html

Repos (ranked 46 of 46 tracked, as of 2026-09-20)

codecrafters-io/build-your-own-x: ranked #1 of 46 by GitHub stars among the entries we track; stars +0.2% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/codecrafters-io-build-your-own-x.html
sindresorhus/awesome: ranked #2 of 46 by GitHub stars among the entries we track; stars +0.5% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/sindresorhus-awesome.html
public-apis/public-apis: ranked #3 of 46 by GitHub stars among the entries we track; stars +0.4% since 2026-09-13 (as of 2026-09-20). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/repos/public-apis-public-apis.html

This week's churn: 1 new entries, 40 not seen in the latest snapshot, 49 rank moves

convaiinnovations/laya moved from #73 to #66 (+53.2% downloads)
XingChen-AGI/Xing4.0-29B-A4B moved from #54 to #51 (+73.4% downloads)
Jackrong/Qwopus3.8-27B-Flash-GGUF moved from #52 to #54 (+0.0% downloads)

Biggest model gains (downloads)

Model	Gain	From → To
`prism-ml/Ternary-Bonsai-2-27B-gguf`	+1,502,787	405,609 → 1,908,396
`Qwen/Qwen3-0.6B`	+1,246,500	22,498,727 → 23,745,227
`cross-encoder/ms-marco-MiniLM-L6-v2`	+977,086	87,728,418 → 88,705,504
`stabilityai/stable-diffusion-xl-base-1.0`	+691,623	2,396,887 → 3,088,510
`ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF`	+447,647	769,557 → 1,217,204

Biggest repo gains (stars)

Repo	Gain	From → To
`browser-use/jev-ultrafast`	+4,970	5,420 → 10,390
`sindresorhus/awesome`	+2,278	505,763 → 508,041
`public-apis/public-apis`	+2,014	479,757 → 481,771
`rakanki911/DLSS5-Swapper`	+1,787	4,206 → 5,993
`eternity4719/HowToLiveBetter`	+1,610	5,024 → 6,634

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-09-20
sindresorhus/awesome — 2026-09-20
public-apis/public-apis — 2026-09-20
freeCodeCamp/freeCodeCamp — 2026-09-20

Hugging Face (343 items)

prism-ml/Ternary-Bonsai-2-27B-gguf — 2026-09-20
deepseek-ai/DeepSeek-V4.1-Flash — 2026-09-20
convaiinnovations/laya — 2026-09-20
XingChen-AGI/Xing4.0-29B-A4B — 2026-09-20

arXiv (431 items)

Coding Agents with an Obstacle-Aware Harness for Safe Robot Manipulation — 2026-09-20
Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision — 2026-09-20
FAMOS: Feed-Forward 3D Articulation Modeling from Sparse Observations — 2026-09-20
Paint-Anything: Unified Any-Color Control for Image Generation and Editing — 2026-09-20

LMSYS Arena (327 items)

claude-fable-5-high — 2026-09-20
claude-opus-4-6-high — 2026-09-20
claude-opus-4-7-high — 2026-09-20
muse-spark-1.2 (xHigh) — 2026-09-20

YouTube (121 items)

Prison system is an insane asylum | Andrew Scull and Lex Fridman — 2026-09-20
Sigmund Freud vs Carl Jung: Why Freud hated Jung | Andrew Scull and Lex Fridman — 2026-09-20
Mental illness in women vs men | Andrew Scull and Lex Fridman — 2026-09-20
The Growing Gap Between ChatGPT and the AI Inside OpenAI - Noam Brown — 2026-09-20

Silicon Valley news (191 items)

Microsoft agentically ports Copilot runtime to Rust for $120K — 2026-09-20
US policy analysts, lawmakers, and others say several state-level AI chatbot safety bills included language that could provide loopholes for tech companies (Katie McQue/NPR) — 2026-09-20
Don't Be Nice — 2026-09-20
Spain Orders Blocks on Archive.today and Its Mirrors — 2026-09-20

Industry news (1053 items)

Nvidia CEO Jensen Huang calls for AI to be developed "as fast as we can" - CBS News — 2026-09-20
Nvidia-backed AI cloud firm Nscale reveals revenue surge in US IPO filing - Reuters — 2026-09-20
Nvidia CEO Jensen Huang emerges as Trump's top ally in AI safety debate - CNBC — 2026-09-20
Dario Says AI Should Slow Down. Jensen Wants to Go Full Steam Ahead. - WSJ — 2026-09-20

AI hardware (175 items)

AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-09-20
AMD Appoints Tim Ryan to Board of Directors — 2026-09-20
AMD Acquires Taalas to Advance Compute Solutions for Rapidly Growing AI Inference Market — 2026-09-20
AMD Reports Second Quarter 2026 Financial Results — 2026-09-20

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

AI signals this week (2026-09-09 2026-09-15): what actually moved

John — Tue, 15 Sep 2026 14:56:16 +0000

Every day I collect AI signals from public sources for Signals 4 — a free daily digest of 4 items, one per core source. This is what the week looked like in the data.

2679 items collected across 7 days and 8 sources.

Biggest model gains (downloads)

Model	Gain	From → To
`cross-encoder/ms-marco-MiniLM-L6-v2`	+1,209,846	86,661,012 → 87,870,858
`Qwen/Qwen3.8-27B`	+991,240	6,712,160 → 7,703,400
`google-t5/t5-small`	+979,691	23,696,426 → 24,676,117
`sentence-transformers/all-MiniLM-L6-v2`	+876,161	253,331,994 → 254,208,155
`unsloth/Qwen3.8-27B-GGUF`	+853,520	10,675,683 → 11,529,203

Biggest repo gains (stars)

Repo	Gain	From → To
`deepseek-ai/deepseek-harness`	+4,309	216,212 → 220,521
`public-apis/public-apis`	+2,820	477,601 → 480,421
`sindresorhus/awesome`	+1,945	504,267 → 506,212
`lnkiai/m3e-canvas`	+1,667	5,201 → 6,868
`vinta/awesome-python`	+1,361	319,380 → 320,741

What showed up, by source

GitHub (140 items)

codecrafters-io/build-your-own-x — 2026-09-15
sindresorhus/awesome — 2026-09-15
public-apis/public-apis — 2026-09-15
freeCodeCamp/freeCodeCamp — 2026-09-15

Hugging Face (342 items)

deepseek-ai/DeepSeek-V4.1-Flash — 2026-09-15
Edge0/Edge0-35B-A3B-preview — 2026-09-15
openbmb/MiniCPM5-2B — 2026-09-15
nex-agi/Nex-N2.5-mini — 2026-09-15

arXiv (282 items)

Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection — 2026-09-15
Stellar Colosseum: A Many-Agent Harness for Long-Horizon Research in Mathematics and Theoretical Computer Science — 2026-09-15
The Router Within: Eliciting Native Skill Routing from a Frozen LLM — 2026-09-15
Vulnerability Localization Benchmark: Measuring Agentic Security Analysis at Repository Scale — 2026-09-15

LMSYS Arena (344 items)

claude-fable-5 — 2026-09-15
claude-opus-4-6-high — 2026-09-15
claude-opus-4-7-high — 2026-09-15
muse-spark-1.2 (xHigh) — 2026-09-15

YouTube (162 items)

Claude Is Now Leaving Invisible Fingerprints In Its Text — 2026-09-15
Why So Many AI Models Sound Like Claude — 2026-09-15
Why Chinese AI Labs Are Keeping Up — 2026-09-15
When will AI be better than human experts? — 2026-09-15

Silicon Valley news (189 items)

Industry news (1045 items)

Trump calls Nvidia CEO to talk AI — and gets put on speakerphone during industry summit - NBC News — 2026-09-15
Samsung backs Nvidia AI chip rival in $230 million funding round as GPU alternatives boom - CNBC — 2026-09-15
Pinterest builds a new foundation for multimodal AI with NVIDIA - Pinterest Newsroom — 2026-09-15
Nvidia Stock Drops. It’s Still the Best Chip Bet in an AI Slowdown. - Barron's — 2026-09-15

AI hardware (175 items)

AMD, Cisco and HUMAIN Expand Saudi Arabia’s AI Infrastructure as AMD Instinct Systems Go Live — 2026-09-15
AMD Appoints Tim Ryan to Board of Directors — 2026-09-15
AMD Acquires Taalas to Advance Compute Solutions for Rapidly Growing AI Inference Market — 2026-09-15
AMD Reports Second Quarter 2026 Financial Results — 2026-09-15

Notes

All links point to the original sources; nothing is rewritten.
The daily digest is free and archived permanently: https://data.jiangzhang.ca/signals4/archive/
Machine-readable summary: https://data.jiangzhang.ca/llms.txt
Raw daily data is also mirrored on GitHub: https://github.com/jiangzhangcc-glitch/ai-signals-daily

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

How to stay current on AI without drowning: a 2-minute daily system

John — Mon, 07 Sep 2026 03:15:58 +0000

How to stay current on AI without drowning: a 2-minute daily system

*Tags: ai, productivity

The AI world changes hourly. The problem was never a lack of information — it's that more information now costs you focus.

Here is the system I use (and the product it turned into).

1. Stop following the firehose

Pick one input per domain, not twenty:

Code trends → GitHub stars/rising
Model landscape → Hugging Face trending
Deep dives → one YouTube channel/programmer you trust
Industry context → one Silicon Valley/industry feed

If an update isn't worth reading in 2 minutes today, it wasn't important yesterday either.

2. Force a "top of the pile" view

For each domain, ask: what's #1 right now? The #1 repo, the #1 model, the freshest relevant video, the biggest story. A "top of the pile" view gives you signal without a ranking obsession.

3. Always keep the original link

Secondhand summaries drift. Keep the ability to jump to the original source — publish time and publisher tell you more than a clickbait headline ever will.

4. Make it daily and automatic — then delete the tabs

Consistency beats intensity. A daily 2-minute pass, automated, replaces hours of anxious tab-opening. If your system requires willpower, it will fail on busy days.

The thing this turned into

I built exactly this as a free product: Signals 4 — one signal each from GitHub, Hugging Face, YouTube and Silicon Valley, four items a day, original links attached, no card required.

See the full product: https://data.jiangzhang.ca/signals4/go?to=landing&src=devto
Subscribe to the free digest: https://data.jiangzhang.ca/signals4/go?to=subscribe&src=devto

What's your system for staying current? Tell me what works for you.

I build **AI Signals* — a free daily digest of AI signals (4 a day, one from each core source)
and an open dataset of everything collected. Read today's edition:
data.jiangzhang.ca/signals4.*

If you need more than the free tier (8 sources, 4 updates a day, full history),
that's the paid feed —
or pay once, CA$5, no subscription.

DEV Community: John

AI signals this week (2026-10-26 to 2026-11-01): what actually moved

The standing numbers (from our own endpoints)

Models (ranked 312 of 2149 tracked, as of 2026-11-01)

Repos (ranked 76 of 76 tracked, as of 2026-11-01)

This week's churn: 4 new entries, 37 not seen in the latest snapshot, 306 rank moves

Biggest model gains (downloads)

Biggest repo gains (stars)

What showed up, by source

GitHub (140 items)

Hugging Face (341 items)

arXiv (434 items)

YouTube (106 items)

Silicon Valley news (189 items)

Industry news (1005 items)

AI hardware (175 items)

Notes

AI signals this week (2026-10-19 to 2026-10-25): what actually moved

The standing numbers (from our own endpoints)

Models (ranked 281 of 1906 tracked, as of 2026-10-25)

Repos (ranked 70 of 70 tracked, as of 2026-10-25)

This week's churn: 1 new entries, 39 not seen in the latest snapshot, 186 rank moves

Biggest model gains (downloads)

Biggest repo gains (stars)

What showed up, by source

GitHub (140 items)

Hugging Face (341 items)

arXiv (446 items)

YouTube (84 items)

Silicon Valley news (193 items)

Industry news (1044 items)

AI hardware (175 items)

Notes

AI signals this week (2026-10-12 to 2026-10-18): what actually moved

The standing numbers (from our own endpoints)

Models (ranked 255 of 1669 tracked, as of 2026-10-18)

Repos (ranked 66 of 66 tracked, as of 2026-10-18)

This week's churn: 5 new entries, 42 not seen in the latest snapshot, 214 rank moves

Biggest model gains (downloads)

Biggest repo gains (stars)

What showed up, by source

GitHub (140 items)

Hugging Face (341 items)

arXiv (423 items)

LMSYS Arena (141 items)

YouTube (117 items)

Silicon Valley news (193 items)

Industry news (1049 items)

AI hardware (175 items)

Notes

44 days of AI signals as a dataset — and 3 silent API traps I hit publishing it

What the dataset is

Trap 1 — Your CDN 403s stdlib clients, and your logs show nothing

Trap 2 — Hugging Face's per-file upload endpoint is gone (410)

Trap 3 — Don't hardcode the namespace

The common thread

43 days, 17,040 AI signals: four things the data says that trend lists won't tell you

1. Only 23 of 1,493 tracked models have enough history to compute a trend

2. Top entities are boringly stable — the interesting signal is churn

3. Most of the volume is one source, and it isn't the interesting one

4. Derive numbers from raw snapshots, or don't publish them

What I'm doing with this

Building an MCP server for AI signals data — and the Cloudflare 403 that silently broke it

1. The MCP server (single file, stdlib only, no API key)

2. The trap: a 403 that your server never sees

The data behind it

AI signals this week (2026-10-05 to 2026-10-11): what actually moved

The standing numbers (from our own endpoints)

Models (ranked 219 of 1426 tracked, as of 2026-10-11)

Repos (ranked 61 of 61 tracked, as of 2026-10-11)

This week's churn: 2 new entries, 37 not seen in the latest snapshot, 183 rank moves

Biggest model gains (downloads)

Biggest repo gains (stars)

What showed up, by source

GitHub (140 items)

Hugging Face (340 items)

arXiv (439 items)

LMSYS Arena (327 items)

YouTube (69 items)

Silicon Valley news (193 items)