# CandleAI quad-check — market, UX, team, next plan

**Updated:** 2026-08-04  
**Purpose:** Second-pass “quad” review after message-integrity + dark-instrument ship. Improve the living plan without rewriting history.  
**Lanes used:** (1) Market/competitors · (2) Product/UX · (3) Team/sales enablement · (4) Ops/QC/AI capacity  
**AI assist:** Spark Qwen3-coder critique OK; Gemma MCP still 401 after auth — human+Cursor arbiter for trust merge.

---

## Quad verdict (one screen)

| Lane | Status after redesign | Biggest remaining gap |
|------|----------------------|------------------------|
| **Market** | Honest limits + dual CTA match ProteinIQ/Tamarind *pattern* | Not yet *feeling* like minutes-to-value; SEO thin |
| **UX** | Dark instrument, no GIF, science nav | Free-run → result still ops-heavy; friendliness lag |
| **Team** | Master plan + training exist | Sales pack not one click; roles ≠ weekly cadence |
| **Quality/ops** | QC-A/QC-F, hub status, www=doctrine | Job UX polish; Trust Center; Spark smoke ritual |

**Positioning that still wins:**  
> ColabFold is free for one. CandleAI is managed structure jobs — queue, QC, reports on dedicated Spark — research-use only.  
> We sell **less babysitting + accountable delivery**, not “another 250-model hub.”

---

## 1) Who to learn from (named players)

### A. Free path (the real competitor for many labs)

| Player | What they are | Steal | Ignore / don’t copy yet |
|--------|---------------|-------|-------------------------|
| **ColabFold / Google Colab notebooks** | DIY free fold | Honesty: free exists; our wedge is ops burden | Competing on “also free AF2” |
| **AlphaFold Server (DeepMind)** | Trusted research web UI | Brand-level clarity of research terms | Pretending we have DeepMind brand |
| **Local installs** | Free weights, your GPU | Pain language (CUDA/MSA/queue) | Becoming an installer vendor |

### B. Self-serve SaaS hubs (UX + freemium bar)

| Player | Signal (2026) | Steal | Don’t copy yet |
|--------|---------------|-------|----------------|
| **[ProteinIQ](https://proteiniq.io/)** | Task→tool→Run; Free ~100 credits / 3 jobs/day; academic vs commercial ladder | Tool landing pages; compare tables; transparent limits | Full multi-tool GPU marketplace |
| **[Benchling.ai / Model Hub](https://www.benchling.com/blog/scientific-models-free-fast-and-for-everyone)** | Sign up → FASTA→prediction in ~1 min; up to ~$1k/yr free framing; calm chrome | Calm hierarchy; “no sales call to start”; credit metaphor | Requiring their ELN ecosystem |
| **[Tamarind Bio](https://www.tamarind.bio/)** | Outcome verticals + free start + Book demo + [Trust Center](https://trust.tamarind.bio/) + API | Dual CTA; security narrative; outcome language | Claiming 100+ / 250 live tools |
| **OpenProtein.AI** | Freemium + private cloud + “own your IP” | IP / data residency story (matches our cyber strength) | Over-claiming design platforms |

### C. Enterprise / sales-led (how money really closes)

| Player | Motion | Steal for CandleAI team |
|--------|--------|-------------------------|
| **Tamarind Premium / Enterprise** | Book meeting; SSO/SLA | When Forge/BATCH: security one-pager + call |
| **Schrödinger et al.** | Procurement-heavy pilots | QC-F + disclaimer pack as “compliance theater lite” |
| **Basecamp / EDEN-class** | Narrative + distribution | Not our lane — don’t chase |

### D. Where CandleAI uniquely sits

| Axis | CandleAI today | Closest peer pattern |
|------|----------------|----------------------|
| Product shape | Focused ColabFold hub + managed ladder | ProteinIQ *depth* without breadth |
| Funnel | Self-serve **1×400** + managed **≤3** + paid | Benchling free start + Tamarind dual CTA |
| Trust | QC-A/QC-F before invoice; research disclaimer | Stronger gate than most free hubs |
| Infra story | Single DGX Spark, pause-vLLM coexistence | Honest scarcity vs “infinite B200 farm” |
| Team edge | Cyber + SWE + PO + geo sales (US/EU/MENA) | Can sell **security + delivery**, not just buttons |

---

## 2) UX / user-friendliness gaps (ranked)

Ship fixed trust/copy. Friendliness is the next crisis if ignored.

| Rank | Gap | Why labs bounce | Fix (iteration) |
|------|-----|-----------------|-----------------|
| P0 | **Post-submit fog** on free run | “Did it work?” unclear vs ProteinIQ job UI | Status timeline: queued → folding → QC → download; email optional |
| P0 | **Example sequence one-click** | Empty FASTA intimidates | “Try demo GFP fragment” button (shared research seq) |
| P1 | **Result packaging** | PDB + JSON feels engineer-native | One “Lab report” card: mean pLDDT, PASS/FAIL, download ZIP |
| P1 | **Mobile form density** | Pilot form long | Progressive disclosure: contact first, FASTA optional later |
| P1 | **Language bridge** | Too much GPU/ops jargon | PO+Elias glossary: lab words first, Spark words secondary |
| P2 | **Trust Center stub** | Commercial buyers ask security | 1-page: encryption, retention, no-PHI default, staff access — cyber owners |
| P2 | **Progress / ETA** | Queue idle vs busy opaque | Hub strip already exists — surface ETA band (“often < N min when idle”) |
| P3 | **Account / history** | Return visits restart | Later; don’t block MVP |

**Friendliness principle:** every public screen answers three questions in <10 seconds — *What can I run? What’s free? What do I get back?*

---

## 3) Team → motivation + information (enablement matrix)

People sell when they have **one packet**, **one talk track**, and **one weekly win**.

| Role / people | Strength | Give them | Weekly ask |
|---------------|----------|-----------|------------|
| **Christian** | Platform / arbiter | Frozen doctrine + release board | Approve copy/deploy; unblock Spark |
| **Ali** | Ops | Intake triage checklist | Pilot inbox → status labels |
| **Elias** | Docs / timeline | Disclaimer + glossary + training refresh | 1 UX copy pass / week |
| **Christian** | Platform / deploy | Trust Center + deploy survivability | www smoke after every deploy |
| **Derek** | Folds / Spark | Smoke ritual + QC samples | Weekly Live smoke; queue health |
| **Pierre** | DACH connections | DE/EN one-pager + talk track | 3 warm intros / month (approval drafts) |
| **Maria / Etienne** | MENA presence | Same pack localized tone | Warm intros; no send without gate |
| **PO / SWE (collective)** | Product + quality | UX backlog above; Playwright-style checks when ready | Ship one friendliness item / sprint |
| **Sales-ready individuals** | Relationships | **Sales Motivation Pack** (below) | Log outcomes in tracker; celebrate first QC-A win |

### Sales Motivation Pack (single folder / staff page)

1. **30-second talk track** (pain → free proof → QC → ask)  
2. **Exact limits card** (1×400 / ≤3×400 / paid after QC-F)  
3. **Proof link** (1GFL QC-A + `/proof.html`)  
4. **Who we are not** (not clinical; not 250 tools)  
5. **Competitor cheat sheet** (§1 table, one page)  
6. **Objection cards:** “ColabFold is free” / “Why not ProteinIQ?” / “Where’s SOC2?”  
7. **Outreach drafts** stay `awaiting_approval` — motivation ≠ auto-send  

Wire into: `staff/master-plan.html` + short `training/06-sales-pack.html` (next build).

---

## 4) How the plan improves (vs redesign GTM plan)

| Prior plan phase | Keep | Upgrade now |
|------------------|------|-------------|
| Phase 0 message integrity | Done on www | Automate prod copy gate in release cycle |
| Phase 1 design system | Dark instrument shipped | **Friendliness sprint:** demo seq, job timeline, lab report card |
| Phase 2 money path | Ladder correct | Add **priority queue SKU story** when wait times hurt; keep BATCH/Forge |
| Phase 3 staff plan | Exists | Add Sales Pack + geo owners + weekly cadence |
| Phase 4 AI lanes | Documented | Fix Gemma MCP; Qwen drafts for SEO/UX copy only |
| Phase 5 test gates | QC-F + snapshot | Add **UX acceptance** (3 questions) + weekly Spark smoke checklist |

### 30-day iteration themes (parallel tracks)

1. **Friendliness** — demo FASTA, job status UX, lab report card  
2. **Market learning** — publish/refresh SEO vs DIY + AF2 when-to-use; watch ProteinIQ/Tamarind pricing pages monthly  
3. **Trust** — Trust Center stub (cyber); www bind ritual already in CI  
4. **Sales motion** — pack + 5 approved outreach targets; first managed pilot with QC-A evidence  
5. **Quality** — weekly fold smoke; no Live label without it  

### Explicit non-goals (still)

Tamarind-scale tool count · clinical claims · auto-send email · full MSA download · Cursor-only marketing drafts when Qwen/Gemma can draft.

---

## 5) Offering optimization (use our advantages)

| Advantage | Product/GTM use |
|-----------|-----------------|
| **Cybersecurity talent** | Trust Center, retention, no-PHI defaults, staff access model — sell to startups that fear SaaS black boxes |
| **Software engineers** | Job UX, API later, release cycle, observability of queue |
| **Product owners** | Friendliness backlog, glossary, acceptance checklist ownership |
| **Geo / relationship sellers** | Warm intros with pack; regional academic + biotech; human SLA is a feature |
| **1× Spark honesty** | Scarcity → fair-use free + paid priority later (don’t fake infinite GPU) |

**Wedge line for sellers:**  
“ProteinIQ/Benchling are great when you want a marketplace credit. CandleAI is for labs that want a **managed fold with a human QC story** on dedicated hardware — start free on one sequence.”

---

## 6) Acceptance checklist (next ship)

- [x] Free run: demo sequence button works  
- [x] Free run: status states readable without staff (timeline)  
- [x] Result: single lab-report summary above raw artifacts  
- [x] Staff: Sales Motivation Pack linked from Master Plan  
- [x] Trust Center stub linked from footer (non-clinical)  
- [ ] Weekly Spark smoke logged in tracker (ops ritual)  
- [x] Prod copy gate command (`message-integrity`) — automate in CI local files  


---

## Related SoR

- Lab hub: `candleai-lab-hub-v1-p0.md`  
- Competitive notes: `candleai-competitive-pilot-marketing-p0.md`  
- GTM money: `candleai-gtm-profitability-p0.md`  
- Staff plan: `candleai-staff-master-plan-p0.md`  
- UI prompts: `candleai-ui-improvement-prompt-framework.md`