Anthropic Ipo Pulse avatar

Anthropic Ipo Pulse

Pricing

Pay per usage

Go to Apify Store
Anthropic Ipo Pulse

Anthropic Ipo Pulse

Anthropic IPO Pulse is an orchestrator Actor that tracks how the internet talks about the Anthropic (Nasdaq: ANT) IPO. It collects posts from X, LinkedIn, YouTube, Threads and stock news. Each post is scored by an LLM for sentiment, relevance and how extreme the take is.

Pricing

Pay per usage

Rating

0.0

(0)

Developer

Fabian Maume

Fabian Maume

Maintained by Community

Actor stats

0

Bookmarked

1

Total users

1

Monthly active users

4 hours ago

Last modified

Categories

Features

Limited permissions

Share

Anthropic IPO Pulse is an orchestrator Actor that tracks how the internet talks about the Anthropic (Nasdaq: ANT) IPO. It collects posts from X, LinkedIn, YouTube, Threads and stock news. Each post is scored by an LLM for sentiment, relevance, how extreme the take is, and its topic. The Actor saves daily aggregates and the most over-optimistic and over-pessimistic quotes to a named key-value store, where a website can read them.

What it does

StepActorInterface
X search, day by day (four 6-hour slices) or the live windowxquik/x-tweet-scraperActor run
LinkedIn postsharvestapi/linkedin-post-searchActor run
YouTube videosstreamers/youtube-scraperActor run
Threads posts (keyword search)lergassy/threads-scraperActor run
Stock newspublicmoney/marketwatch-scraper, publicmoney/yahoo-finance-scraper (news mode)Standby HTTP
Quotespublicmoney/yahoo-finance-scraper, publicmoney/nasdaq-scraperStandby HTTP
Ticker news sentimentpublicmoney/sentiment-miniStandby HTTP
Post scoringapify/openrouterStandby HTTP

Modes

  • backfill: collects every day from dateFrom to dateTo and writes them to the baseline. Run it once.
  • daily: collects yesterday. Schedule it at 00:15 UTC so LinkedIn's "newest first" results cover the day that just ended.
  • news: sweeps MarketWatch and Yahoo Finance news only. Schedule it every 2 hours, because the feeds keep only about a day of stories.
  • rescore: no scraping. Re-scores posts already saved in <storeName>-posts for the date range and rebuilds those days, storylines included. Use it when a run hit its LLM budget, or once to add topics to posts scored before topics existed (they are sent to the LLM again).
  • live: snapshots the last liveWindowMinutes (default 15) for X and Threads, plus the news sweep and quotes. Schedule it every 15 minutes on IPO day. LinkedIn and YouTube stay daily-only.

Sentiment sampling

The LLM scores up to maxScoredPerPlatformDay (default 300) posts per platform per UTC day: the most-engaged third, so the quote wall gets the strongest takes, plus an even spread of the rest. Volume counts every collected post; the relevant count is volume × the relevance rate of the scored sample. News is always scored in full.

Topics and storylines

The scoring prompt also tags every relevant post with one topic:

TopicCovers
valuationprice, valuation, IPO pricing, first-day pop
bubbleAI bubble or overvaluation fears
governanceLong-Term Benefit Trust, founder control, public benefit corporation structure, board
safetyAI safety and Anthropic's mission
accessretail access, allocations, how and where to buy
rivalscomparisons with OpenAI, SpaceX, Nvidia or other listings
businessClaude products, revenue, customers, losses, compute
insiderslockups, employee and investor selling, early backers
macrorates, markets, Nasdaq mood
othernone of these

After backfill, daily and rescore runs write a day's aggregates, the Actor makes one more LLM call per day. It sends up to 40 of the day's most-engaged relevant scored posts and asks for at most 3 storylines, each citing 1 to 4 of those posts. Links the LLM did not get are dropped, and so is a storyline left without a link. Days with fewer than 8 relevant scored posts are skipped. If the call fails, the day keeps its previous storylines and the run carries on.

Output

The named key-value store (default anthropic-ipo) holds these records:

  • baseline: per UTC day and per platform, the volume, relevant post count, engagement, sentiment sums, bull/bear/neutral counts, the top quotes on each side, and whether a page cap was hit. It also holds the daily quote and ticker-sentiment snapshots. Each day and platform also carries:
    • topics: per topic, n (judged relevant posts with a score) and sentimentSum (the sum of their scores). Divide to get the topic's average sentiment.
    • extremeBull and extremeBear: judged relevant posts with an extremeness of 0.7 or more, by stance.
    • storylines: per UTC day, up to 3 objects with title (70 characters at most), summary (one sentence, 180 characters at most), stance, topic and urls (1 to 4 source posts).
  • live and live-timeline: the latest live window and its history. Timeline entries carry each platform's topics too.
  • costs: the estimated Apify spend per step and per run.
  • meta: the queries, model and Actors used.

Every scored post goes to the run's dataset and to the named dataset <storeName>-posts, with its topic (null when the post was not scored or is not about Anthropic).

Cost control

Set Max total charge (USD) on the run. The budget is split evenly across the seven steps (X, LinkedIn, YouTube, Threads, news, markets, LLM). Each batch Actor run gets its share as maxTotalChargeUsd, and each Standby step stops once its estimated spend reaches its share. A daily run with default settings costs about $1, so give it a cap of about $3: the even split leaves each step about $0.43, enough for LinkedIn's roughly $0.30. A live snapshot costs about $0.25 (cap $1), so a 12-hour IPO day at 4 snapshots an hour is about $12.

Storylines add one LLM call per day, about $0.005 with the default model, so a daily run costs under a cent more. The topic field adds a few output tokens per scored post. A rescore that adds topics to posts scored before topics existed re-scores up to maxScoredPerPlatformDay posts per platform per day, about $0.10 per day of history. It only spends on the LLM step, which gets one eighth of the cap, so give it a cap of about 8 times the LLM spend you expect.

Requirements

The account must be able to use apify/openrouter (it works on the free plan). If scoring fails, posts are still counted for volume but carry no sentiment.

License

MIT