CEREBRO: Statistics

What each channel fetched, and what survived

All-time totals from the 72 of 78 briefings that carry a per-channel ledger; the earliest briefings predate the block and are skipped. Channel spellings are folded, so a channel that reports under two names in the raw notes appears here once. Days is how many of those briefings the channel appeared in, the honest stand-in for a time axis on an all-time table.

Channel	Fetched	In briefing	Keep rate	Days
rss	9,620	538	5.6%	72
x	4,847	248	5.1%	72
hackernews	3,810	183	4.8%	72
github	3,388	360	10.6%	72
show_hn	2,800	17	0.6%	72
reddit	2,200	236	10.7%	72
yc-rfs	972	0	0.0%	72
github_search	340	5	1.5%	17
yc-launch	339	8	2.4%	72
gmail	333	5	1.5%	72
swipe	303	11	3.6%	17
seed_urls	17	1	5.9%	17
total	28,969	1,612	5.6%	72

Trends

Signals per briefing

Jun 19 to Sep 6, 2026. Busiest day June 19 with 25; the floor is a day nothing cleared the bar.

Score distribution

Buckets of 0.05, with the bounds taken from the corpus rather than from the model’s nominal range: published scores run 0.50 to 0.95. Anything under the triage bar never reaches a briefing, which is why the left edge is where it is.

0.50–0.55: 19
0.55–0.60: 87
0.60–0.65: 10
0.65–0.70: 381
0.70–0.75: 22
0.75–0.80: 217
0.80–0.85: 420
0.85–0.90: 5
0.90–0.95: 91

Category mix, September 2026

coding-agents-llm: 78 · 61.4%
agentic-saas: 34 · 26.8%
vibe-coding: 11 · 8.7%
cli-tui: 4 · 3.1%

Category share over time

Month	Signals	coding-agents-llm	agentic-saas	vibe-coding	cli-tui
June 2026	216	146 · 67.6%	49 · 22.7%	18 · 8.3%	3 · 1.4%
July 2026	410	316 · 77.1%	52 · 12.7%	33 · 8.0%	9 · 2.2%
August 2026	499	317 · 63.5%	102 · 20.4%	55 · 11.0%	25 · 5.0%
September 2026	127	78 · 61.4%	34 · 26.8%	11 · 8.7%	4 · 3.1%

Where the links go

Hostnames across all 1,252 signal URLs, 118 distinct. The top 25 are listed; 93 further hosts appear at least once. These are deliberately plain text rather than links: this page is indexed, and publishing ninety outbound hrefs here would rebuild exactly the outbound link graph the archive is noindex to avoid. The links themselves live on the briefings, where they belong.

github.com269
reddit.com258
x.com206
producthunt.com199
github.blog70
latent.space41
simonwillison.net39
interconnects.ai13
arxiv.org11
openai.com11
swipe.md8
news.ycombinator.com5
anthropic.com4
huggingface.co3
kimi.com3
z.ai3
artificialanalysis.ai2
blog.cloudflare.com2
blog.modelcontextprotocol.io2
cacm.acm.org2
claude.com2
danluu.com2
primeintellect.ai2
12gramsofcarbon.com1
acceptmarkdown.com1

Colophon · What it costs