Hello World

Outbound is broken.

Not because the tools don't work. Because everyone is using the same three tools — Apollo, ZoomInfo, Outreach — scraping the same database, firing at the same 30,000 accounts, opening with the same four templates. The reader of a cold email today has seen the exact same email from forty other companies this quarter.

The signal-to-noise ratio collapsed. Reply rates followed. A 1.5% reply rate is now considered good.

What we built

Black Magic is a hybrid GTM agent platform. An open-source Electron desktop app runs six GTM workflows locally against your own vault — a folder of markdown files that describe who you sell to, what you sell, how you talk, and which signals mean "now is a good time." A remote backend at blackmagic.engineering proxies every upstream call (LLM, research, enrichment, LinkedIn) so you never hold API keys and credits are metered centrally.

Five agents ship in the dashboard:

• Website Visitors — deanonymizes inbound traffic, qualifies against your ICP, pushes to your CRM.
• LinkedIn Intent — scrapes LinkedIn content / jobs / title changes, matches to your triggers, drafts sequences.
• Meeting Prep — pre-meeting account intelligence in under 60 seconds.
• Lookalike Outbound — takes a closed-won customer, finds firmographic + behavioral lookalikes.
• Closed-Lost Analysis — replays old CRM losses against new signals, surfaces which should be re-opened.

None of these are "an AI BDR." They are narrow, composable agents that do one job each. You review the output. You ship the output.

Why "signal-based"

The contact database is dead as a moat. When everyone buys the same list, the list is worth zero. The moat is the signal layer — the live events that turn a cold account into a warm one:

• Hired 3+ AEs in the last 90 days.
• Raised a Series A or B in the last 45 days.
• New VP Sales / CRO in seat.
• Public complaint about a tool they're paying for.
• Quiet change in their tech stack.

Signals come from places most outbound teams don't scrape — LinkedIn jobs feeds, Reddit pain posts, G2 review corpora, Crunchbase firehoses, podcast guest lists. Not Apollo. Not ZoomInfo. The whole point is to not be where everyone else is.

Why vault-native

Every agent reads from ~/BlackMagic/us/ on every turn — one directory of markdown files describing your business. ICP, voice, positioning, past wins, triggers, tools. Markdown, not prompts. Version-controlled, not hidden in a SaaS setting pane.

The tools disappear when the context is right. The context lives in files you own.

Change a file, and every subsequent draft changes. Move a customer win into the vault, and the agent starts citing it in comparable pitches. Rewrite your voice guidelines, and cold emails inherit the new tone the same afternoon.

Why open-source desktop

Because your company data should not live in a SaaS we can cut off tomorrow. The vault is a folder on your machine. The desktop app is MIT. The remote backend is the credit meter and the upstream-API proxy — nothing else. If we ever go away, you walk away with every byte of context you built.

What comes next

We are not building a "full-stack marketing platform." We are not replacing your CRM, your sequencer, or your inbox. We sit on top of what you already run — HubSpot or Salesforce below us, Instantly or Smartlead above us — and turn signals into sequences.

Every week, we ship a new trigger, a new data source, a new sequence angle. The whole system is designed to be extended from the outside: a markdown file in us/triggers.md, a scraper in scrapers/, a sequence variant in us/sequences/.

This is the starting line. The system will get sharper every quarter.

---

This is our hello to the world that is coming. Download the desktop app at blackmagic.engineering. Read the architecture at github.com/furudo-erika/blackmagic-ai.