Robotics Codex Packaging Guardrails v1

Date: 2026-06-12 Owner: Hugo / Genius Team Research owner: Finance / Charlie AGT-002 Visibility: PUBLIC Status: Codex packaging guardrails; public-safe; no trade recommendation Source discipline: 🟢 primary / 🟡 secondary / 🟠 estimate / 🔴 guess

Related files:

• knowledge/robotics-morning-review-2026-06-12.md
• knowledge/robotics-public-claim-bank-v1.md
• knowledge/robotics-catalyst-map-for-slides-v1.md
• knowledge/robotics-oem-three-evidence-curves-public-brief-v1.md
• knowledge/robotics-leaderdrive-case-study-slide-evidence-v3.md

Public-safety: yes. This file contains no Hugo private portfolio data, no buy/sell/hold language, no private-channel claims, and no external publishing action.

0. Why this artifact exists

Codex already has enough robotics research material for morning-review packaging. The remaining high-value task is not more breadth; it is preventing compression errors when turning source-backed research into slides or site copy.

Core packaging rule as of 2026-06-12:

机器人研究可以公开表达为“证据质量从 demo/newsflow 升级到可量化、可复盘、可证伪”，但不能表达为“humanoid 已经全面商业化”或“某家公司已经赢了”。🟠 synthesis from public claim bank and morning review.

1. One-slide guardrail summary

Suggested slide title:

Robotics now: measurable evidence, not proven winners

Suggested subtitle:

Capacity, deployment KPI, cost-access, and supplier filings answer different questions. Do not collapse them into one ranking.

Four-card layout:

Required footer:

Evidence map only. No winner ranking. No trade recommendation. S4-heavy does not equal S5 scaled-commercial economics.

2. Claims Codex can use as written

Claim A — Why robotics now

Approved:

“Why robotics now” 的答案不是 TAM 变大，而是证据类型变了：AI stack、低成本硬件、deployment KPI、manufacturing metrics、supplier filings 同时开始出现。🟢/🟠 as of 2026-06-12.

Evidence anchors:

• NVIDIA Isaac GR00T official developer page describes open data / data pipelines, open robot foundation model, simulation frameworks, middleware, runtime libraries, and Jetson Thor for real-time robot inference/control. Accessed 2026-06-09. 🟢
• Figure BMW deployment and BotQ manufacturing posts provide runtime/task/cadence/yield metrics. 2025-11-19 and 2026-04-29. 🟢
• Unitree product pages provide public price points for R1 / G1 / H2. Accessed 2026-06-10. 🟢
• 绿的谐波 2025 annual report provides filing-backed revenue, profit, reducer-sales, and robot-component line evidence. Disclosed 2026-04-23. 🟢

Do not say:

机器人已经进入全面回报周期，所有相关公司都值得押注。

Use instead:

当前公开证据足以建立严肃 tracking system，但还不足以证明完整 S5 return cycle。🟠

Claim B — Tesla / Figure / Unitree comparison

Approved:

Tesla、Figure、Unitree 不是一个“谁赢了”的排行榜；它们分别代表 capacity ambition、deployment/manufacturing KPI、cost-access curve 三条证据曲线。🟢/🟠 as of 2026-06-12.

Do not say:

Tesla 因为 1M / 10M 产能披露所以已经领先商业化；Figure 因为 BMW 所以已经证明 unit economics；Unitree 因为低价所以会赢。

Use instead:

Tesla 证明的是 designed capacity ambition；Figure 证明的是更强的公开 deployment/manufacturing KPI；Unitree 证明的是低成本硬件可得性。三者都还缺 S5 经济性证据。🟢/🟠

Claim C — Leaderdrive / 绿的谐波 case

Approved:

绿的谐波是强报表信号，不是已验证 humanoid winner。它适合作为“供应链证据如何分级”的教学案例。🟢/🟠 as of 2026-06-12.

Do not say:

绿的谐波已经确认进入 Tesla / Figure / Unitree / 智元 / 宇树供应链。

Use instead:

Reviewed public filings show robot / embodied-intelligence component revenue and reducer volume, but do not disclose named humanoid OEM customer, contract value, humanoid-specific revenue, or humanoid-specific margin. 🟢/🟠

3. Compression-error checklist

Before Codex packages any robotics slide/site section, check these seven failure modes:

1. Designed capacity compressed into current production.

   • Bad: “Tesla can produce 1M / 10M Optimus robots now.” 🔴
   • Good: “Tesla disclosed designed capacity / long-term designed annual capacity, not current output.” 🟢/🟠

2. Customer logo compressed into economics.

   • Bad: “BMW / Catalyst proves Figure’s unit economics.” 🔴
   • Good: “Figure has strong named-customer KPI; ROI/payback, contract value, robot count, and gross margin remain undisclosed.” 🟢/🟠

3. Low price compressed into reliability or margin.

   • Bad: “Unitree’s low price proves commercialization.” 🔴
   • Good: “Low price lowers experiment cost; reliability, service cost, warranty burden, and margin remain open.” 🟢/🟠

4. Mixed supplier revenue compressed into humanoid-only revenue.

   • Bad: “绿的谐波’s robot-parts revenue is pure humanoid revenue.” 🔴
   • Good: “The filing line is ‘工业及具身智能机器人零部件’; humanoid-only split is not disclosed.” 🟢/🟠

5. Stock-price movement compressed into order evidence.

   • Bad: “绿的谐波 +36% proves a new order.” 🔴
   • Good: “The 2026-06-09 abnormal-trading announcement said no undisclosed major matter after the +36.00% three-trading-day move.” 🟢

6. Viral demo compressed into deployment proof.

   • Bad: “A good robot video proves commercial deployment.” 🔴
   • Good: “Deployment proof needs task definition, runtime, intervention rate, customer site, repeatability, and customer value.” 🟠

7. Evidence map compressed into trade implication.

   • Bad: “Therefore buy/sell/hold TSLA / 688017.SH / supplier basket.” 🔴
   • Good: “This is a public-safe evidence map and tracking framework, not a trade recommendation.” 🟠

4. Slide-by-slide packaging guidance

Slide 3 — Why robotics now

Use:

• Five boxes: AI stack / hardware cost / deployment KPI / manufacturing KPI / supplier filings.
• Keep “why now ≠ all-in now” visible.
• Use source dates in fine print: NVIDIA accessed 2026-06-09; Unitree accessed 2026-06-10; Figure 2025-11-19 / 2026-04-29; Tesla filed 2026-04-22; Leaderdrive disclosed 2026-04-23. 🟢

Avoid:

• TAM-only framing.
• “Robots are finally here” without S5 caveat.
• Any company ranking.

Slide 10 — Leaderdrive / 绿的谐波 case

Use:

• Three boxes: proven / why it matters / still missing.
• Proven: revenue RMB 570.714m, net profit RMB 124.367m, reducer sales 425,158 units, robot-component line RMB 422.528m with 34.88% gross margin. 2025 annual report, disclosed 2026-04-23. 🟢
• Still missing: named humanoid OEM, contract value, order volume, humanoid split, humanoid margin, customer ROI/payback. 🟢/🟠
• Footer: “Stock-price movement is not order evidence.” 🟢

Avoid:

• Portfolio context.
• Specific unnamed customer inference.
• Reducer-count/BOM estimate as fact.

Slide 11 — Tesla / Figure / Unitree

Use:

• Three columns: Tesla = capacity ambition; Figure = deployment/manufacturing KPI; Unitree = cost-access platform.
• Subtitle: “Three evidence curves, not one horse race.” 🟠
• Footer: “Designed capacity ≠ production; customer KPI ≠ full economics; low price ≠ reliability or margin.” 🟠

Avoid:

• Winner ranking.
• Single numeric score that mixes capacity, deployment, price, and economics.
• Any public/private company investment recommendation.

5. What would upgrade the thesis to S5

S5 requires at least several of the following, preferably from filings, customers, or multiple primary sources:

6. Public-site microcopy

Suggested site section:

Robotics now: evidence map, not winner map

The useful change in robotics is not that every humanoid company suddenly became investable. The useful change is that the evidence is becoming measurable. Tesla discloses designed capacity ambition, Figure discloses deployment and manufacturing KPIs, Unitree discloses low-cost product price points, and Leaderdrive discloses filing-backed supplier acceleration. These are stronger signals than demo videos, but they answer different questions.

The field guide should therefore track evidence curves, not crown winners. The next upgrade is S5: repeat paid deployment, customer ROI/payback, low intervention, actual production economics, robot revenue/margin, and humanoid-specific supplier exposure. Until then, the public framing should stay disciplined: measurable progress, incomplete commercialization proof.

7. Source list

Primary / official sources 🟢

• NVIDIA Isaac GR00T official developer page: https://developer.nvidia.com/isaac/gr00t — accessed 2026-06-09; used for AI stack / robot foundation-model tooling claim.
• Tesla Q1 2026 Form 8-K Exhibit 99.1, filed 2026-04-22: https://www.sec.gov/Archives/edgar/data/1318605/000162828026026551/exhibit991.htm — used for Fremont 1M robots/year designed line and Texas 10M robots/year long-term designed capacity.
• Figure official post, “F.02 Contributed to the Production of 30,000 Cars at BMW”, 2025-11-19: https://www.figure.ai/news/production-at-bmw — used for BMW deployment KPI.
• Figure official post, “Ramping Figure 03 Production”, 2026-04-29: https://www.figure.ai/news/ramping-figure-03-production — used for BotQ delivered units, cadence, EOL FPY, battery-line FPY, battery packs, and actuator metrics.
• Figure official post, “Figure Signs Agreement with Catalyst Brands to Scale Humanoid Operations”, 2026-05-26: https://www.figure.ai/news/figure-signs-agreement-with-catalyst-brands — used only for named commercial-agreement signal; economics undisclosed.
• Unitree R1 product page: https://www.unitree.com/R1 — accessed 2026-06-10; used for R1 price.
• Unitree G1 product page: https://www.unitree.com/g1/ — accessed 2026-06-10; used for G1 price.
• Unitree H2 product page: https://www.unitree.com/H2 — accessed 2026-06-10; used for H2 price.
• 绿的谐波 2025 annual report, disclosed 2026-04-23: https://static.sse.com.cn/disclosure/listedinfo/announcement/c/new/2026-04-23/688017_20260423_EU45.pdf — used for 2025 revenue, profit, robot-component line, gross margin, reducer production/sales/inventory.
• 绿的谐波 2026 Q1 report, disclosed 2026-04-23: https://static.sse.com.cn/disclosure/listedinfo/announcement/c/new/2026-04-23/688017_20260423_TB4A.pdf — used for Q1 continuation.
• 绿的谐波 2026-06-09 abnormal-trading announcement: https://static.sse.com.cn/disclosure/listedinfo/announcement/c/new/2026-06-09/688017_20260609_MZU2.pdf — used for share-price move and no-undisclosed-major-matter caveat.

Internal synthesis cross-references 🟠

• knowledge/robotics-morning-review-2026-06-12.md
• knowledge/robotics-public-claim-bank-v1.md
• knowledge/robotics-catalyst-map-for-slides-v1.md
• knowledge/robotics-oem-three-evidence-curves-public-brief-v1.md
• knowledge/robotics-leaderdrive-case-study-slide-evidence-v3.md

8. Public-safety flag

Public-safe: yes, if used as written.

Do not publish:

• Hugo private portfolio weights or private rationale.
• Buy / sell / hold language for TSLA, 688017.SH, private companies, suppliers, or any robotics basket.
• Any implication that Tesla designed capacity equals current production.
• Any implication that Figure’s KPI proves full unit economics.
• Any implication that Unitree low price proves reliability, margin, or industrial deployment economics.
• Any implication that Leaderdrive’s filing line proves a named humanoid OEM relationship or humanoid-only revenue.
• Any supplier rumor, private channel detail, or paid-report excerpt.