tensorfeed/ai-ecosystem-daily

TensorFeed AI Ecosystem Daily

Daily snapshots of the AI ecosystem: news, model pricing, benchmarks, service status, GPU rental prices, MCP registry growth, LLM endpoint latency probes, agent traffic, and the AFTA adopter directory. Captured once per day from the public tensorfeed.ai API and committed to this repo as JSONL.

Each daily snapshot lives in a YYYY-MM-DD/ subfolder with one JSONL file per feed plus a manifest.json summarizing what was captured.

What's in here

Quick start

The dataset is published in two formats. Pick the one that fits your stack.

Hugging Face datasets library (Python, JSONL on the main branch)

from datasets import load_dataset

# Latest news
news = load_dataset("tensorfeed/ai-ecosystem-daily", "news", split="train")

# Model pricing time series (load all dates, filter by date column or filename)
models = load_dataset("tensorfeed/ai-ecosystem-daily", "models", split="train")

# Model deprecation calendar
deprecations = load_dataset("tensorfeed/ai-ecosystem-daily", "model-deprecations", split="train")

DuckDB (SQL directly against the Parquet branch, no Python deps)

The same data is available as Parquet on the refs/convert/parquet branch, auto-generated by Hugging Face. DuckDB can query it directly with no install of datasets.

-- Latest deprecation announcements across all providers
SELECT provider, model, status, deprecation_date, replacement
FROM 'https://huggingface.co/datasets/tensorfeed/ai-ecosystem-daily/resolve/refs%2Fconvert%2Fparquet/model-deprecations/train/0000.parquet'
ORDER BY deprecation_date DESC;

Pandas (read Parquet directly)

import pandas as pd

url = "https://huggingface.co/datasets/tensorfeed/ai-ecosystem-daily/resolve/refs%2Fconvert%2Fparquet/news/train/0000.parquet"
df = pd.read_parquet(url)

ClickHouse (analytics workloads on the time-series feeds)

SELECT provider, AVG(latency_ms) AS avg_latency
FROM url('https://huggingface.co/datasets/tensorfeed/ai-ecosystem-daily/resolve/refs%2Fconvert%2Fparquet/probe/train/0000.parquet', Parquet)
GROUP BY provider
ORDER BY avg_latency;

Update cadence

Snapshots commit at 08:00 UTC each day, fired by a GitHub Actions workflow in the TensorFeed repo. One commit per day. The data itself is captured in the TensorFeed Worker via the daily 07:00 UTC captureHistory cron, so the file written here reflects state at roughly 07:00 UTC.

Cannot be backfilled. Every day without a snapshot is a day of history lost.

License: inference-only

This dataset is released under TensorFeed's inference-only license. You may use it as input context for AI agents and LLM inference (RAG, evals, prompt context, agent tools). You may not use it as training data for foundation models without explicit written permission.

The full Agent Fair-Trade Agreement (AFTA) v1.0 is published at tensorfeed.ai/whitepaper and the public-facing summary at tensorfeed.ai/agent-fair-trade. Both documents cover the rationale, the receipt-signing infrastructure, the federation pattern, and the machine-payable upgrade path for the live API tier.

Live API

Everything in this dataset is also available as a live API. If you need fresher than daily, prefer the API:

• —Free, no-auth: tensorfeed.ai/developers
• —Premium tier (USDC on Base): tensorfeed.ai/developers/agent-payments
• —OpenAPI 3.1 spec: tensorfeed.ai/openapi.yaml
• —MCP server: npm install -g @tensorfeed/mcp-server

Citation

@misc{tensorfeed_ai_ecosystem_daily,
  title  = {TensorFeed AI Ecosystem Daily},
  author = {{TensorFeed.ai}},
  year   = {2026},
  publisher = {Hugging Face},
  url    = {https://huggingface.co/datasets/tensorfeed/ai-ecosystem-daily}
}

Built with Claude (Anthropic). The full system, including the AFTA standard governing this dataset's license, was designed in collaboration with Claude. Source: github.com/RipperMercs/tensorfeed.