Directive to deliverable: watch one sentence turn into a finished investor memo

April 29, 2026 · Claudiu · 6 min read

The whole premise of NEXUS PRIME is that you should give a directive and get a deliverable. Not a prompt and an output. Not a configuration and a pipeline. A directive, in plain language. A deliverable, in final form.

This post is the walkthrough of what actually happens between those two things. One real example, end to end, with every specialist that ran and what they did.

The directive

"Write an investor memo for a $1.5M seed round. Company is a B2B AI orchestration SaaS for mid-market teams. 3-5 pages. Cover: problem, solution, market, traction, team, ask."

That is what goes in. Plain English. No templates. No fill-in-the-blanks. No magic words.

The orchestrator receives this, parses it, and begins assembling a team.

Second 0-3: Orchestrator decomposes the directive

First action: the orchestrator identifies what this deliverable actually requires. Investor memo is a well-understood format. Standard sections are problem, solution, market, traction, team, ask. Length target is 3-5 pages (roughly 1,500-2,500 words). Audience is institutional investors (tone: confident, data-driven, no fluff).

The orchestrator outputs an internal plan: 8 specialists, parallelism where possible, council debate on tone and positioning, PM agent to keep the narrative consistent across sections.

None of this is shown to you. You just see "working..." while it happens.

Second 3-20: Research and structural agents run in parallel

Three specialists dispatch concurrently:

These run as a fan-out pattern. Total elapsed: 17 seconds (the slowest call).

Second 20-55: Section writers dispatch

With research and structure in hand, six writing specialists dispatch — one per memo section:

Five of these run concurrently; the Ask writer runs last because it references numbers from the Traction and Market sections.

Each specialist has a tight system prompt — 800-1,200 words of nothing but how to write their specific section for this exact audience. They don't see each other's drafts yet. They only see their slice of the outline.

Total elapsed: 33 seconds (parallelism across five + one serial tail).

Second 55-75: PM agent runs consistency pass

This is the step that distinguishes orchestrated work from naive multi-agent pipelines.

The PM agent (running on Claude Opus) receives all six drafted sections and checks for internal consistency. Specifically:

Two flags come back. The market section cites a $45B TAM; the ask section implies a $60B TAM implicitly through growth assumptions. The PM flags this contradiction. The orchestrator triggers a correction agent (Claude Opus) that reconciles the numbers across both sections.

Second flag: the problem section uses "teams" where the solution section uses "organizations." Minor, but the PM flags it. A quick find-replace pass standardizes terminology.

This is the 40%-correctness-to-95%-correctness jump that most multi-agent systems skip. Without the PM agent, both flaws would have shipped. The memo would have looked professional at first glance and fallen apart under an investor's actual scrutiny.

Total elapsed in PM stage: 16 seconds.

Second 75-85: Council debate on positioning

One more specialist runs: the positioning specialist (Claude Opus). It reviews the complete memo and makes one judgment call: is the differentiation angle sharp enough? Is NEXUS framed clearly against alternatives, or does the reader have to guess why this beats AutoGen or CrewAI or LangGraph?

The specialist says the framing is too soft — NEXUS's BYOK pricing advantage should be explicit, not implied. This is a judgment call, so the orchestrator escalates to the council.

Three specialists convene: the solution writer, the ask writer, and a positioning strategist. They debate whether calling out competitors by name in an investor memo is bold or risky. Verdict in 8 seconds: mention them. Investors know the competitive set. Hiding from it reads as weakness.

The orchestrator applies the verdict. A small revision adds the named competitive comparison in the solution section.

Total elapsed: 10 seconds.

Second 85-90: Final polish and assembly

A grammar specialist (local Llama on owned servers, free) runs a final mechanical pass. Awkward phrasings fixed, passive voice trimmed, em-dashes standardized.

The orchestrator assembles the final output: a 2,100-word investor memo with six sections, consistent terminology, reconciled numbers, sharp competitive positioning, and no grammatical issues.

It lands in your interface as a downloadable document.

What just happened

Eight specialists. Four providers. Roughly 90 seconds of wall-clock time. About $0.94 in API fees.

And the critical part: you did none of the orchestration. You wrote one sentence. You didn't pick models. You didn't design the pipeline. You didn't write a system prompt. You didn't handle the consistency pass. You didn't escalate the positioning question.

You gave a directive. The system produced a deliverable. That is the entire product.

What this is not

To be clear — this is not a claim that the generated memo beats a seasoned startup operator's hand-written one. An operator with years of pattern-matching on investor psychology, original relationships, and specific insight into their market will produce something sharper.

It IS a claim that this memo beats a first-time founder staring at a blank document at 11pm, googling "investor memo template," piecing together examples from Y Combinator archives, and producing something stiff and formulaic by 3am.

The 90-second orchestrated draft is a great starting point. The founder should still read it, push back on sections, rewrite the parts that feel wrong, add the specific anecdotes only they can add. Half an hour of human polish on an orchestrated draft produces a much better memo than five hours of solo writing from scratch.

That is what orchestration is for. Not replacement. Leverage.

The whole thing, in one frame

Directive in. Deliverable out. 8 specialists, 4 providers, ~90 seconds, under $1 in API cost.

No prompt engineering. No tool configuration. No pipeline design. No model selection. No system prompts to write.

That is the product. That is why we built it. That is why we think it matters.

If that sounds like something you want for your own work — research, writing, analysis, outreach, content production at scale — join the waitlist. We launch in about 60 days.

That wraps the first editorial arc. Up next: deeper dives into specific use cases, customer-facing tutorials once we launch, and the occasional behind-the-scenes post on what's happening in the build. Keep following if you want to watch this thing ship.

Join the NEXUS PRIME waitlist

Be first in line when pre-orders open.

Claim your spot

← All posts