AI News Digest
Fetch and curate important AI news from the user's X/Twitter AI list.
Prerequisites
@dannote/bird-premium package (official X API v2)
- AI list ID:
1894700501725229467
Workflow
- Fetch tweets from the last 24 hours (use pagination to get all)
- Filter out noise (marketing, challenges, hiring, retweets, emojis-only)
- Categorize into: Releases, Papers, Insights, Tools
- Extract actual announcement links from tweet text (t.co URLs to blogs, GitHub, arxiv)
- Include tweet link as source
Fetching Last 24 Hours
The list produces ~250-300 tweets per day. Use pagination to fetch all:
# First request (100 tweets max per request)
bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --json
# Continue with cursor from previous response until createdAt < 24h ago
bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --cursor <nextCursor> --json
Stop fetching when the oldest tweet's createdAt is older than 24 hours.
Merge all tweets from paginated responses, then filter by createdAt >= (now - 24h).
Filtering Criteria
Include (important)
- Model releases and announcements
- Research papers (arxiv links, paper titles)
- Tool/library releases with technical details
- Benchmark results
- Technical insights and analysis
- Infrastructure updates (vLLM, ComfyUI, MLX, etc.)
- API launches
Exclude (noise)
- Marketing fluff ("excited to announce", "join us")
- Challenges and contests
- Hiring posts
- Retweets (text starting with "RT @")
- Single emoji or short reaction posts
- Partnership announcements without technical substance
- Repetitive product links spam
Output Format
Organize into sections. Extract actual links from tweet text when available (GitHub, arxiv, blog posts). Include tweet as source.
## Releases
**ProductName** (Company) — Brief description
Link: https://github.com/org/repo (or blog/arxiv link from tweet)
Source: https://x.com/username/status/ID
## Papers
**Paper Title** — One-line summary (Institution)
Link: https://arxiv.org/abs/XXXX
Source: https://x.com/username/status/ID
## Insights
**Topic** — Key finding or leak
Source: https://x.com/username/status/ID
## Tools
**Tool** — What it does
Link: https://example.com
Source: https://x.com/username/status/ID
Example Usage
User: "What's happening in AI today?"
- Fetch first page:
bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --json
- Check oldest tweet's
createdAt — if < 24h ago, fetch next page with --cursor
- Repeat until oldest tweet is > 24h old
- Merge all tweets, filter to last 24h by
createdAt
- Parse JSON, extract:
author.username, text, id
- Build source link:
https://x.com/{username}/status/{id}
- Extract t.co links from text, resolve to actual URLs when relevant
- Filter noise, categorize, output curated digest
1---
2name: ai-news
3description: Fetch and curate AI news from X/Twitter list. Use when user asks for AI news, AI digest, what's happening in AI, or wants a summary of AI announcements, papers, and releases.
4metadata:
5 source: https://skillmd.com/skills/dannote/ai-news
6---
7
8# AI News Digest
9
10Fetch and curate important AI news from the user's X/Twitter AI list.
11
12## Prerequisites
13
14- `@dannote/bird-premium` package (official X API v2)
15- AI list ID: `1894700501725229467`
16
17## Workflow
18
191. Fetch tweets from the last 24 hours (use pagination to get all)
202. Filter out noise (marketing, challenges, hiring, retweets, emojis-only)
213. Categorize into: Releases, Papers, Insights, Tools
224. Extract actual announcement links from tweet text (t.co URLs to blogs, GitHub, arxiv)
235. Include tweet link as source
24
25## Fetching Last 24 Hours
26
27The list produces ~250-300 tweets per day. Use pagination to fetch all:
28
29```bash
30# First request (100 tweets max per request)
31bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --json
32
33# Continue with cursor from previous response until createdAt < 24h ago
34bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --cursor <nextCursor> --json
35```
36
37**Stop fetching when** the oldest tweet's `createdAt` is older than 24 hours.
38
39**Merge all tweets** from paginated responses, then filter by `createdAt` >= (now - 24h).
40
41## Filtering Criteria
42
43### Include (important)
44
45- Model releases and announcements
46- Research papers (arxiv links, paper titles)
47- Tool/library releases with technical details
48- Benchmark results
49- Technical insights and analysis
50- Infrastructure updates (vLLM, ComfyUI, MLX, etc.)
51- API launches
52
53### Exclude (noise)
54
55- Marketing fluff ("excited to announce", "join us")
56- Challenges and contests
57- Hiring posts
58- Retweets (text starting with "RT @")
59- Single emoji or short reaction posts
60- Partnership announcements without technical substance
61- Repetitive product links spam
62
63## Output Format
64
65Organize into sections. Extract actual links from tweet text when available (GitHub, arxiv, blog posts). Include tweet as source.
66
67```markdown
68## Releases
69
70**ProductName** (Company) — Brief description
71Link: https://github.com/org/repo (or blog/arxiv link from tweet)
72Source: https://x.com/username/status/ID
73
74## Papers
75
76**Paper Title** — One-line summary (Institution)
77Link: https://arxiv.org/abs/XXXX
78Source: https://x.com/username/status/ID
79
80## Insights
81
82**Topic** — Key finding or leak
83Source: https://x.com/username/status/ID
84
85## Tools
86
87**Tool** — What it does
88Link: https://example.com
89Source: https://x.com/username/status/ID
90```
91
92## Example Usage
93
94User: "What's happening in AI today?"
95
961. Fetch first page:
97 ```bash
98 bunx --bun @dannote/bird-premium list-timeline 1894700501725229467 -n 100 --json
99 ```
1002. Check oldest tweet's `createdAt` — if < 24h ago, fetch next page with `--cursor`
1013. Repeat until oldest tweet is > 24h old
1024. Merge all tweets, filter to last 24h by `createdAt`
1035. Parse JSON, extract: `author.username`, `text`, `id`
1046. Build source link: `https://x.com/{username}/status/{id}`
1057. Extract t.co links from text, resolve to actual URLs when relevant
1068. Filter noise, categorize, output curated digest