Faro Dev Update: Two Months of Building Trust

For the past two months, we put trust ahead of new features. First, we rebuilt the data foundation. Then we made the agent and alerts more reliable on top of it.

Everything we shipped rests on that work: alerts that fire on multiple conditions, and answers that come from verified data.

Here's what we rebuilt, what shipped on top of it, and what's coming next.

1. We rebuilt the data foundation

We learned something: when a metric was calculated ahead of time, the agent's answers held up. So we rebuilt the foundation before building anything else on top of it.

Before writing new code, we audited 101 metrics across 6 data families.

What changed?

→ 621 data tests check every change before it ships

→ 20,000+ rows that our data sources deleted are now kept permanently

→ 0 missing or out-of-range values served

→ One definition per metric

→ 1,725 field descriptions with full data lineage, generated automatically

→ "As-of" tables that answer "what did we know on day X?", which makes honest backtesting possible

The new pipeline is running in production and verified: 50 serving tables, 27 cleaned-data models, and 30 history snapshots archiving every version of every row since September 24.

2. A retrieval harness for the agent

We built a retrieval harness: a layer that finds the right data for each question before the agent answers. Soon, we'll connect it to the agent so it can use semantic search to find the right metric for each question.

Charting is also built and wired into the agent, widgets, and dashboards.

3. A sturdier agent in the making

→ Reliability: Fixes to cancellation and connections, plus automatic retries for temporary model errors.

→ Compaction: Long chats now continue in the same thread.

→ Stop and continue: Stop a reply mid-stream and keep going in the same chat.

→ Branching: Copy a whole conversation, or branch from a single message, to explore another scenario without losing the original.

→ Image input: Drop in a chart screenshot or a screenshot of an X post, ask "analyze this", then keep asking follow-ups.

→ Personalization: Answers adapt to your experience level, preferred length, analysis focus, and time horizon.

4. Alerts you can rely on: Composite Alerts are live

Composite Alerts let you combine multiple conditions into one alert.

→ Describe it or build it: Tell Faro what you want ("BTC above 84,500 and funding rate crosses above 13%, notify me once, push"), and a dedicated alert agent sets it up. Review the settings and confirm, or build it yourself in the builder.

→ Choose data you trust: Pick from strategies like Cowen Band Cross, Bollinger Squeeze, 2-Year MA Multiplier and more, alongside datasets like price, funding rate, RSI, EMA, liquidations, positioning and more.

→ Telegram delivery: Alerts now reach you on Telegram as well as by push.

→ Chart deep links: Tap an alert and it opens at that exact point on the chart.

5. Public signals: Faro posts to X on its own

The same alert engine now publishes public signals. Set conditions, pick a card template, run a backtest, and publish to Slack or X. Get a sneak peek on @HLRekt.

6. The HYPE and BTC data dashboards are live

HYPE and BTC now have the most comprehensive research pages available, right here on Faro.

What's next?

→ Agent on the new foundation: We're switching the agent to the dictionary and the new tables, adding more metrics and question types, and expanding to ETH and SOL.

→ Portfolio V2: This brings Faro to casual investors, not just active traders.

• Your trades analyzed: win rate, average win and loss, and what trading cost you

• Winners and losers: which assets made and lost you the most

• Your portfolio over time, with Faro explaining what's driving it

→ Strategies: A way to set your rules and have Faro execute while you're away. It's in the works. More soon.

Thanks for building with us. Keep the feedback coming ⚒️