Skip to content

Commit abf6aec

Browse files
spaarke-devspaarke-dev
authored
Merge pull request #53 from spaarke-dev/content/context-is-only-one-layer
content: context-is-only-one-layer -> review (NetDocuments PR counter, LinkedIn)
2 parents 7f8b4aa + 09f42c9 commit abf6aec

7 files changed

Lines changed: 449 additions & 0 deletions

File tree

‎content-platform/articles/context-is-only-one-layer/CLAUDE.md‎

Lines changed: 51 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,51 @@
1+
# Per-piece session contract — context-is-only-one-layer
2+
3+
> Session router for a LinkedIn post. Load files in this order before
4+
> producing any prose. Generated by the `content-pipeline` skill.
5+
6+
## 1. The constitution (always)
7+
8+
Read parent `../../CLAUDE.md` first. The non-negotiable load order:
9+
10+
1. `../../voice/style-guide.md`
11+
2. `../../voice/brand-positioning.md`
12+
3. `../../voice/audience-personas.md`
13+
4. `../../content-types/linkedin-post.md`
14+
15+
## 2. This piece (always)
16+
17+
1. `brief.md` — the spec for THIS post
18+
2. `plan.md` — hook + body beats + close (do not draft until signed off)
19+
3. `tasks.md` — workflow gates
20+
21+
## 3. Selective references (load on demand)
22+
23+
- `../../voice/product-knowledge.md` — when touching Spaarke architecture
24+
- `../../voice/domain-knowledge.md` — when touching industry stats
25+
- `../../voice/vocabulary.md` — consult when reaching for a word
26+
- `../../voice/examples/good-articles.md` — for syndication tone
27+
- `../../voice/examples/avoid-this.md` — **always run final-pass**
28+
29+
## 4. Adjacent context
30+
31+
- **Campaign:** `brief.md` `campaign:` field → `../../campaigns/<slug>.md`
32+
- **GitHub Issue:** `brief.md` `github_issue:` field
33+
34+
## 5. Workflow
35+
36+
1. **Outline** — fill plan.md, get hook + structure signed off
37+
2. **Draft** — write to `draft.md` in this directory
38+
3. **Revise** — apply reviewer feedback in same file
39+
4. **Polish** — hook reads cold, hashtags ≤3, byline correct
40+
5. **Image** *(optional)* — carousel or syndication header
41+
6. **Ship** — move to `published/linkedin-posts/`, manual post to LinkedIn
42+
43+
See tasks.md for the done-when checklist per gate.
44+
45+
## 6. What you don't do here
46+
47+
- Do not auto-post to LinkedIn. Posting is human-driven.
48+
- Do not exceed 3 hashtags.
49+
- Do not write a hook that requires the writer to be already known
50+
to the reader.
51+
- Do not skip the outline gate.

‎content-platform/articles/context-is-only-one-layer/brief.md‎

Lines changed: 87 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,87 @@
1+
---
2+
slug: context-is-only-one-layer
3+
type: linkedin-post
4+
publish_date: 2026-05-18
5+
channels: [linkedin]
6+
status: brief # brief | outline | draft | review | scheduled | published
7+
priority: high # time-sensitive: NetDocuments PR landed 2026-05-14
8+
audience: corporate-counsel # primary; legal-ops-director secondary
9+
length_target: 280 # words for standalone (150–400)
10+
byline: ralph-schroeder # founder voice; personal LinkedIn profile
11+
campaign: 2026-05-spaarke-launch # fits launch arc + market-timing hook
12+
triggered_by: NetDocuments "Context Graph for Legal Work" press release, 2026-05-14
13+
hook: |
14+
A context graph is necessary but not sufficient.
15+
The Legal IQ platform also requires inference and memory.
16+
format: standalone # standalone | carousel | syndication
17+
hashtags: # 2 — at the bottom on their own line
18+
- LegalOps
19+
- LegalTech
20+
---
21+
22+
# Topic
23+
24+
A subtle counter-positioning post in response to a recent industry announcement that "context graphs" are the next big legal-AI category. The take: a context graph is one architectural layer — the operating model is the broader frame. Names the Legal IQ stack (Data, Memory, Inference) explicitly. Does not name the competitor.
25+
26+
# Angle / Point of view
27+
28+
Spaarke has been building the operating-model argument from day one; the market is now arriving at part of it. The post agrees with the premise that context matters and institutional knowledge is the prize, then broadens the frame: context graph alone is a memory-layer artifact. What lawyers actually need is the full operating layer — Data + Memory + Inference grounded in their own tenant, with AI directing and humans controlling. The walkthrough is the proof; the existing articles carry the long form.
29+
30+
The tone is founder voice — confident, not defensive, not aggressive. Anyone in the field will know which announcement triggered this. That's the point.
31+
32+
# Why now
33+
34+
NetDocuments published "NetDocuments Unveils the First Context Graph for Legal Work" on 2026-05-14. Foundation Capital and Gartner have begun framing "context engineering" as the strategic priority for enterprise AI. Spaarke is already in this category and ahead on the architecture — the broader frame (operating model > graph) is worth planting before the field crowds with similar framing. Founder voice the next-business-day after a competitor's category move is the right cadence: timely without looking reactive.
35+
36+
This post also reinforces the [2026-05 Spaarke Launch](../../campaigns/2026-05-spaarke-launch.md) campaign by adding a third moment after Welcome (May 11) and the foundation-piece re-promotions — without bumping any of the planned May 19 / May 20 pieces.
37+
38+
# Must include
39+
40+
- The hook line verbatim (locked in frontmatter)
41+
- The Legal IQ stack named explicitly: Data, Memory, Inference. Used **structurally** to organize the argument, not just name-checked (per `voice/style-guide.md` §5.10).
42+
- The "operating model" frame as the broader claim — what makes a context graph operational rather than ornamental
43+
- A subtle reference to the engagement boundary (in-house + outside counsel on the same record) — the structural difference even when both sides have a graph
44+
- CTA: invitation to **see the Spaarke platform walkthrough** as the closing move — show, not tell
45+
- Two link-outs at the bottom of the post:
46+
- [The Legal IQ Stack: Data, Memory, Inference](/why-spaarke/the-iq-stack) — primary architectural tie-in
47+
- [Institutional Knowledge Is Walking Out the Door](/why-spaarke/institutional-knowledge) — the same problem they're naming, our long form
48+
- Hashtags `#LegalOps #LegalTech` at the bottom on their own line
49+
50+
# Must NOT include
51+
52+
- The name "NetDocuments" or any competitor name. The post argues the category, not the announcement.
53+
- A feature list of Spaarke vs anyone else. The frame is architectural, not feature-by-feature.
54+
- Generic "we have AI too" framing. The take is specific: context graph is one layer, operating model is the frame.
55+
- Marketing closes ("Schedule a demo"), exclamation points, or empty intensifiers — per `voice/style-guide.md` §5.
56+
- Organizational voice. Byline is `ralph-schroeder`; sustain first-person throughout.
57+
- A 600-word post in tight paragraphs with H2-style structure (per `content-types/linkedin-post.md` §9 — that's the "blog voice on LinkedIn" pitfall).
58+
- The phrase "Spaarke AI has three pillars" or anything referencing the old three-pillar framing. The current architecture is **three core layers** — Foundry IQ for grounding and operational memory, Microsoft Agent Framework for orchestration and execution, Microsoft Power Platform and Microsoft 365 for user experience. (See `voice/product-knowledge.md` §3.)
59+
60+
# References
61+
62+
- [The Legal IQ Stack: Data, Memory, Inference](/why-spaarke/the-iq-stack) — primary architectural tie-in; the post organizes around its three layers
63+
- [Institutional Knowledge Is Walking Out the Door](/why-spaarke/institutional-knowledge) — the operational-memory case for what walks out the door when senior staff leave; the "institutionalize what was in lawyers' heads" problem the trigger PR names
64+
- [Welcome to Spaarke](/blog/welcome-to-spaarke) — the launch piece this post sits inside the campaign of (may not need to link, but reinforces the campaign arc)
65+
- [Why We Built on Microsoft](/why-spaarke/why-we-built-on-microsoft) — the tenant-inside vs tenant-adjacent architectural case (use **only** if we want to make the architecture point explicitly; my recommendation is to hold this for the Day-2 company-page echo, not this post)
66+
- **The Spaarke platform walkthrough** — the closing CTA. URL: https://www.spaarke.com/semantic-search-walkthrough
67+
68+
# Voice notes
69+
70+
Sustained first-person from `ralph-schroeder`. Founder cadence — confident, plainspoken, operator-grade. Short sentences. White space between paragraphs (mobile preview is the rendering target). One earned superlative maximum; preferably zero. The hook is the most load-bearing line; it has been pre-approved verbatim and must not be rewritten.
71+
72+
# Hero graphic
73+
74+
Optional. Default: **skip** — founder text posts often land harder cold than with a visual, and the post's claim is the asset.
75+
76+
If a visual is added, reuse `/brand/diagrams/spaarke-ai-architecture.svg` (the updated three-layer architecture diagram) at 1:1 or 4:5 portrait. Alt text: "Spaarke's three core architectural layers: Foundry IQ for grounding and operational memory, Microsoft Agent Framework for orchestration and execution, Microsoft Power Platform and Microsoft 365 for user experience."
77+
78+
# Follow-up
79+
80+
A separate **Spaarke company-page echo post** is planned for 1–2 business days after this one (target 2026-05-20). Lighter framing, organizational byline, different opening — picks up the operating-model thread for the corporate-counsel feed that doesn't follow Ralph's profile. Tracked separately; do not write that here.
81+
82+
---
83+
84+
## Unresolved (resolve before publish)
85+
86+
- [x] **Platform walkthrough URL** — locked: https://www.spaarke.com/semantic-search-walkthrough
87+
- [ ] **Ralph LinkedIn URL on file** — `voice/bylines.md` has it as `**TBD — confirm with team**`. Not blocking for this post; worth confirming in the byline registry separately.

‎content-platform/articles/context-is-only-one-layer/draft.md‎

Lines changed: 37 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,37 @@
1+
---
2+
slug: context-is-only-one-layer
3+
type: linkedin-post
4+
publish_date: 2026-05-18
5+
channels: [linkedin]
6+
status: review # ready for human review
7+
byline: ralph-schroeder
8+
audience: corporate-counsel
9+
length_target: 280
10+
campaign: 2026-05-spaarke-launch
11+
hook: |
12+
A context graph is necessary but not sufficient.
13+
The Legal IQ platform also requires inference and memory.
14+
hashtags: [LegalOps, LegalTech]
15+
---
16+
17+
A context graph is necessary but not sufficient.
18+
The Legal IQ platform also requires inference and memory.
19+
20+
The field is converging on something important: context — what a system knows about a matter, a counterparty, a history — is the prize. Many of this week's announcements get that part right.
21+
22+
But a graph of documents and people is one architectural layer. Legal IQ runs in three. Data: the unified operational record. Memory: institutional history and the connections between matters, counsel, outcomes, decisions. Inference: AI directed at the grounded record, with humans accountable for the call.
23+
24+
A context graph built on a document system sees documents and parties. An operating model sees matters, projects, invoices, spend, counsel performance, OCG compliance, and outcomes — on the same record, with inference applied across them.
25+
26+
And when the platform runs inside the customer's own Microsoft environment — which is how we designed Spaarke from the start — the engagement boundary stops fragmenting AI grounding. The firm and the in-house team work against the same matter, the same memory.
27+
28+
A context graph is necessary. It is not the destination. An operating model is.
29+
30+
If you want to see what that looks like in practice, take the walkthrough:
31+
https://www.spaarke.com/semantic-search-walkthrough
32+
33+
The longer arguments:
34+
https://www.spaarke.com/why-spaarke/the-iq-stack
35+
https://www.spaarke.com/why-spaarke/institutional-knowledge
36+
37+
#LegalOps #LegalTech

‎content-platform/articles/context-is-only-one-layer/idea.md‎

Lines changed: 110 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,110 @@
1+
# Context is only one layer
2+
3+
<!--
4+
slug: context-is-only-one-layer
5+
proposed type: linkedin-post (Ralph's personal profile)
6+
follow-up: Spaarke company-page echo post (separate piece, 1-2 days later)
7+
campaign: 2026-05-spaarke-launch
8+
triggered_by: NetDocuments "Context Graph for Legal Work" press release, 2026-05-14
9+
-->
10+
11+
## The trigger
12+
13+
NetDocuments announced its "Context Graph for Legal Work" on 2026-05-14. They
14+
position it as the first context graph in legal — a graph that maps connections
15+
across hundreds of millions of records, surfaces matter overviews on open,
16+
supports semantic search across the firm, and lets external AI (Claude, ChatGPT)
17+
work against the firm's institutional knowledge via MCP. Built on AWS, integrates
18+
with M365 via MCP. 25+ year DMS history.
19+
20+
Press release: https://www.netdocuments.com/company-news/netdocuments-unveils-context-graph-legal-platform/
21+
22+
## The argument
23+
24+
A context graph is necessary but not sufficient. What lawyers actually need is
25+
an operating model — and the graph is one layer in it (the Memory layer in the
26+
[Legal IQ Stack](/why-spaarke/the-iq-stack) framing).
27+
28+
Three subtle counter-points to weave in, without naming NetDocuments:
29+
30+
1. **A graph built on a DMS sees documents, parties, and timelines. An operating
31+
model sees matters, projects, invoices, spend, counsel performance, OCG
32+
compliance, and outcomes — all on the same record.** Different scope of
33+
memory; different decisions it can inform.
34+
35+
2. **Integrating *with* Microsoft 365 via MCP is not the same as running *inside*
36+
it.** Tenant-adjacent (AWS + M365 connectors) versus tenant-inside (Power
37+
Platform + Dataverse + SharePoint Embedded in the customer's own M365 tenant)
38+
is a structural difference, not a deployment preference. IT security review
39+
treats them differently. Data sovereignty model is different.
40+
41+
3. **A firm's graph and a company's graph aren't connected.** A law firm using a
42+
DMS-based graph and an in-house team using their own still work from separate
43+
tenants on the same matter. Spaarke's three-stakeholder model means the same
44+
matter record and operational memory span the engagement boundary.
45+
46+
## Tone
47+
48+
Subtle. Don't name NetDocuments. Agree with the premise (context matters,
49+
institutional knowledge is the prize), broaden the frame (operating model >
50+
graph). Anyone in the field will recognize the response; that's the point.
51+
52+
Founder voice for the personal post. Not corporate. Not defensive. Confident but
53+
not aggressive.
54+
55+
## Asset and CTA
56+
57+
The post should close with a CTA inviting the reader to **see the Spaarke
58+
platform walkthrough** — show, don't tell. The walkthrough is the differentiator
59+
because it makes the operating-model argument visible in 60 seconds rather than
60+
4 paragraphs.
61+
62+
## Tie-ins (links to include in the post or follow-up comment)
63+
64+
Use 2-3 of these, not all:
65+
66+
- [The Legal IQ Stack: Data, Memory, Inference](/why-spaarke/the-iq-stack) —
67+
primary architectural tie-in; explicitly maps "Memory" as one of three layers
68+
- [Institutional Knowledge Is Walking Out the Door](/why-spaarke/institutional-knowledge) —
69+
NetDocuments names exactly this problem; we have the existing article
70+
- [What is Legal Operations Intelligence?](/why-spaarke/what-is-legal-operations-intelligence) —
71+
the category piece, operating-model frame
72+
- [Why We Built on Microsoft](/why-spaarke/why-we-built-on-microsoft) —
73+
architectural counter on tenant-inside vs tenant-adjacent (use if we want to
74+
make the architecture point explicitly)
75+
- [Welcome to Spaarke](/blog/welcome-to-spaarke) — the launch article; natural
76+
primary cross-link since this sits in the launch campaign
77+
78+
My pick: **IQ Stack** (the architecture answer), **Institutional Knowledge**
79+
(the same problem they're naming), and **the platform walkthrough** (the CTA).
80+
81+
## Hook ideas (draft territory — the brief should lock one)
82+
83+
- "Glad to see the field validating context graphs as a category." [pivot to:
84+
context is one layer; the operating model is the bigger frame]
85+
- "A context graph is necessary but not sufficient." [direct take]
86+
- "Every legal team is about to hear the phrase 'context graph' a lot." [meta
87+
observation — invites the reader in]
88+
89+
## Audience
90+
91+
Primary: corporate-counsel + legal-operations (the buyers who'll see this PR
92+
and weigh it). Secondary: outside-counsel leadership (firm operations leaders
93+
deciding their next-gen DMS).
94+
95+
## What this should NOT become
96+
97+
- An attack on NetDocuments. We don't name them.
98+
- A laundry list of Spaarke features. The point is the architectural argument,
99+
not a feature comparison.
100+
- A general "we have AI too" post. The take is specific: context graph is one
101+
layer, operating model is the frame.
102+
- A long-form blog post. This is a LinkedIn post — punchy, founder voice,
103+
150-400 words.
104+
105+
## Follow-up: Spaarke company-page echo (1-2 days later)
106+
107+
Separate piece, separate slug (e.g., `legal-iq-is-an-operating-model` or
108+
similar). Spaarke org voice. Lighter framing. Same architectural argument,
109+
different opening — targets the corporate-counsel feed that doesn't follow
110+
Ralph's profile. Decide that one after this one lands.

0 commit comments

Comments
 (0)