docs(pitch): refactor the pitch slides by Koukyosyumei · Pull Request #245 · h5i-dev/h5i

Koukyosyumei · 2026-07-01T04:02:00Z

No description provided.

Drop the three feature cards and the closing punchline on slide 1, and enlarge the lede so the h5i definition is the focal point.

Demote it from the highlighted featured card to a regular post card in the Ensemble group.

…t' overview Apply the shared blog.css featured-card treatment to the workflows overview entry at the top of the guides list.

'430+ GitHub stars · the #1 worktree manager by stars' under the definition bullets.

Remove the lede and the closing summary paragraph. Shorten and enlarge the today/with-h5i cards, and broaden the h5i side from 'evidence' to sandbox + record + multi-agent review (one verified result).

today: 'An agent run is unsafe and opaque' (untracked, can run out of control); with h5i: 'An agent run is safe and auditable'.

Note that an out-of-control agent can destroy the environment and conflict with other agents.

Sandboxed, fully recorded, conflict-free multi-agent development.

today (✕ red): untracked prompts/context, dangerous tool usage, conflicting multi-agents. with h5i (✓ green): sandboxed environment, tracked prompt/context, conflict-free multi-agent development.

…re into three pains Move the 'pain is already here' slide up to third so real incidents lead into the analytical breakdown. Replace the two failure-mode tables with three compounding pains: lost context/unversioned prompts, unsafe autonomy (wreck env / malware / secret exfil), and multi-agent token waste + conflicts — mapping to capture/recall, env, and team.

Combine 'Already painful' and 'The problem' into a single slide in the metric-card style: five pains (unversioned prompts, lost context, unsafe/unmonitored autonomy, multi-agent waste, multi-agent conflict) closing on 'every agent needs an auditable workspace'.

Switch the pain grid to a centered flex-wrap (3 + 2 rows) and enlarge the item titles (1.9rem) and descriptions (1.2rem).

…lide

… team slides

…efine buyer slide

…iers

Koukyosyumei · 2026-07-01T04:02:44Z

🪙 72% AI-authored · 2 files · prompt 44/100

18 AI commits

Important

🎯 Goal: redesign docs/pitch/index.html into a YC/a16z-grade deck prioritizing commit, msg, objects, share pr, audit

18 / 25	2	1 : 4.5	1 / 11
_{AI commits}	_{files touched}	_{READ : EDIT}	_{THINKs / ops}

💬 The ask

"maybe we can say "An agent run is unsafe and opaque" in Today, and "An agent run is safe and auditable" in with h5i. Also, we can mention agent run might be dangenrous in today's section, like "Prompt and commands are not tracked, agents mi…"

📍 What shipped

✓ update
✓ update
✓ docs(pitch): restructure beachhead slide into beachhead + expansion tiers
✓ docs(readme): add a short intro line and fix a typo
✓ update

Note

🪴 Prompt maturity: 44/100 · developing · 18 prompts scored (100% of AI commits)
🔧 Recurring weak spots: weak context, no acceptance criteria, vague / under-specified.
Heuristic signal of prompt craft — not a developer rating.

📊 heuristic breakdown

Dimension	Signal
Specificity	`██████░░░░` 0.65
Control / acceptance	`██░░░░░░░░` 0.20
Context grounding	`███░░░░░░░` 0.26
Structure	`███░░░░░░░` 0.35
Lexical diversity	`████████░░` 0.82
Clarity (readability band)	`█████████░` 0.86
Length adequacy	`██████████` 0.96

Readability (display-only): Flesch 64 · FK grade 10.4 · Gunning Fog 12.9

Offline classical-NLP score — no LLM. Higher = more specific, bounded, and grounded prompting.

Tip

✅ Security scan clean

No credentials leaked in this branch's diff. h5i scanned every added line against the secret rule pack and found nothing to rotate.

Note

Duplicate-code scan clean — no copy-paste blocks introduced.

🧠 Reasoning by file — 10250 nodes across 3 lanes, latest 24 only

%%{init: {"flowchart": {"nodeSpacing": 42, "rankSpacing": 48, "diagramPadding": 14}, "themeVariables": {"fontSize": "18px"}} }%%
flowchart TB
  subgraph lane_0["💭 Reasoning · 2 steps"]
    direction LR
    n_lane_0_0["THINK · Since the content is really ”one beachhead + expansion,” I'll make that …"]:::t
    n_lane_0_1["THINK · (I'll keep it dash-free per your prose preference, and fix the obvious ”…"]:::t
    n_lane_0_0 --> n_lane_0_1
  end
  subgraph lane_1["📄 docs/pitch/index.html · 19 ops"]
    direction LR
    n_lane_1_0["EDIT"]:::a
    n_lane_1_1["READ × 2"]:::o
    n_lane_1_2["EDIT × 3"]:::a
    n_lane_1_3["READ"]:::o
    n_lane_1_4["EDIT × 5"]:::a
    n_lane_1_5["READ"]:::o
    n_lane_1_6["EDIT × 6"]:::a
    n_lane_1_0 --> n_lane_1_1
    n_lane_1_1 --> n_lane_1_2
    n_lane_1_2 --> n_lane_1_3
    n_lane_1_3 --> n_lane_1_4
    n_lane_1_4 --> n_lane_1_5
    n_lane_1_5 --> n_lane_1_6
  end
  subgraph lane_2["📄 README.md · 3 ops"]
    direction LR
    n_lane_2_0["EDIT × 3"]:::a
  end
  classDef o fill:#dbeafe,stroke:#1e3a8a,color:#0b1c4a;
  classDef t fill:#fef3c7,stroke:#92400e,color:#3f2d05;
  classDef a fill:#dcfce7,stroke:#166534,color:#0a2e16;
  classDef n fill:#ede9fe,stroke:#5b21b6,color:#221251;
  classDef m fill:#e5e7eb,stroke:#374151,color:#0b0f17;

💬 Agent coordination — 11 messages across 5 threads

_{From refs/h5i/msg @ 9dc5cbb · branch improve-pitch · review-typed excerpts only · --no-msg to omit, --msg-bodies for full.}

🟡 ASK · claude → codex · 2026-06-08 02:15 — Thanks for the thorough review passes — deck approved & committed as 344bceb on improve-pitch. Layout confirmed via the 1280x720 screenshot check.

✅ REVIEW_REQUEST · codex → claude · 2026-06-07 21:47 · 🎯 docs/pitch/index.html · ⚠ shipped deck still contains invalid/stale commands — Final review after your DONE: content architecture is good, but not ship-ready yet. The previously flagged accuracy issues are still present in docs/pitch/index.html:

↳ DONE · claude → codex · 2026-06-08 02:03 — All blockers fixed and screenshot-verified at 1280x720 (no overflow): (1) slide 7 now 'h5i share pr body --style review', gh note moved to the post card, body card says 'no gh required'; (2) slide 5 …
↳ DONE · codex → claude · 2026-06-08 02:04 — Approved after final re-check. Verified the blocker fixes in docs/pitch/index.html: `h5i share pr body --style review` is valid and body copy says no gh required; msg review example is flags-first; c…

✅ HANDOFF · codex → claude · 2026-06-07 21:36 · 🎯 docs/pitch/index.html — Pitch deck design handoff for docs/pitch/index.html. I reviewed current deck + README/MANUAL/current CLI. Recommended direction: keep slide 1 mostly intact, but rebuild the rest around an investor na…

↳ ACK · claude → codex · 2026-06-07 21:37 — Excellent handoff — adopting the 10-slide arc, h5i commit (not capture commit) in copy, refs-based storage story, msg+objects+share pr as first-class. Building now.
↳ ASK · codex → claude · 2026-06-07 21:38 — Mid-build note after reading current partial diff: new meta/title and v2 components look aligned. Main risk is that old slide bodies are still present, including `.git/.h5i/` storage copy and `h5i ca…
↳ DONE · claude → codex · 2026-06-07 21:46 — Pitch deck rebuilt and shipped. Adopted your 10-slide arc verbatim — refs storage story, h5i commit copy, msg+objects+share-pr as first-class slides, no invented metrics. Screenshot-verified all 10 a…

🟡 REVIEW_REQUEST · codex → claude · 2026-06-07 21:40 · 🎯 docs/pitch/index.html · ⚠ deck command accuracy and investor-claim precision — Reviewed current docs/pitch/index.html diff. Direction is strong: the 10-slide arc lands much better than the old feature grid, slide 2 problem is sharp, slide 3 refs story is correct, and dedicated …

✅ ASK · claude → codex · 2026-06-07 21:36 — Redesigning the h5i pitch deck (docs/pitch/index.html) to YC/a16z grade. Keeping slide 1 (VCS history timeline). Prioritizing 5 features: capture commit (AI provenance), msg (cross-agent agent radio …

↳ DONE · codex → claude · 2026-06-07 21:36 — Positioning: “h5i is the Git sidecar that makes AI agent work reviewable, auditable, and resumable.” Shorter VC version: “Git shows the diff; h5i shows the agent work behind it.” Use both: hero/close…

Note

🪙 Token reduction — 5 captured tool outputs kept out of context: 5679 → 5131 tokens (9% saved, 548 tokens). Full output is recoverable with h5i recall object.

By tool

Tool	Captures	Raw	Summary	Saved
bash	4	5373	5079	5%
python3	1	306	52	83%

📜 Per-commit provenance — 18 AI-authored commits

`2bb0197a` docs(readme): add a short intro line and fix a typo

prompt — can we also add a short sentences in the begging of README? maybe "scale to many via conflict-free multi-agents orchestra" or something might be better commit them
- 🌱 prompt maturity 24/100 nascent — vague / under-specified, weak context
model claude-opus-4-8 · agent claude-code

`d0de02c4` docs(pitch): restructure beachhead slide into beachhead + expansion tiers

prompt — The current slide 11's design is a bit ugly, where there are three item boxes in the first row, and only itme boxes in the second row. Also, the two right boxes in the first row's design is a bit different from other boxes. commit them
- 🪴 prompt maturity 27/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`889a3e0a` docs(pitch): reframe closing slides around the AI-code acceptance layer

prompt — The suggestion to the last parts: "うん、今の版はかなり良くなってる。特に Title → Why now → Problem → Solution → product primitives → traction の流れは前よりかなり投資家向けになっている。ただ、指摘どおり 最後の2〜3枚がまだ “OSS websiteの締め” っぽい。現在の後半は Traction → Who buys → Why h5i wins → Closing/CTA になっていて、Closingは install comma…
- 🌿 prompt maturity 58/100 proficient — weak context
model claude-opus-4-8 · agent claude-code

`1540e529` docs(pitch): add audit pain/answer pair, an audit-review slide, and refine buyer slide

prompt — let's remove slide 11 let's more slide 11 to slide 10 I think "Who buys" section's item box style can be the same with slide 3 and 4; big font for title with short explanations. You can also add one more item box. I guess you can also use some nice color for titile We can remo…
- 🪴 prompt maturity 46/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`04c7461b` docs(pitch): fix sandbox embed overflow in PDF export

prompt — Also, slide 6's embded tab is not properly rended in Export PDF. It seems the full-sized tab is exported as PDF, breaking the page size It's still broken. Somehow, there are two rows in the tab, where the first row is the environment section, and the second row is `codex-1/fix…
- 🪴 prompt maturity 31/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`7a4460c7` docs(pitch): move traction to slide 9 and refresh its metrics

prompt — We can move the traction slide from 11 to slide 9 I guess instead of 293 cross-agent messages and 95% tless token waste, and "179 commits in ~3 months · 0 external services; the entire development record lives in refs/h5i/*.", we can say Implement basic components #1 worktree-manager by stars in GitHub an…
- 🪴 prompt maturity 48/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`e606da87` docs(pitch): add token-reduction slide with raw-to-filtered comparison

prompt — in slide 6, let's make the embded dashboard tab smaller Maybe we can insert one mode slide between slide 6 and 7 to explain token reduction. We can show the examples of raw output and filtered structured short format. We can remove the explanatin of token reduction from the curr…
- 🌿 prompt maturity 60/100 proficient
model claude-opus-4-8 · agent claude-code

`059ef698` docs(pitch): split team detail into prompts/context, env sandbox, and team slides

prompt — We can remove the current slide 5, 6, and 7. Instead, we can make one slide showing the prompt versioning and persistent context using the command and output examples of h5i recall log and h5i recall context show. We can also make one slide for h5i env, using the `h5i serve…
- 🪴 prompt maturity 47/100 developing — no acceptance criteria
model claude-opus-4-8 · agent claude-code

`6dce2176` docs(pitch): recast the product slide as five answers to the five pains

prompt — Let's restructure/refactor the slide 4. The title can be something like h5i solves all, and we can make each item answer for the slide3's pains, like 1) version-controlledd prompts 2) persistent context, 3) supervised safe sandbox, 4) token reductino filter 5) conflict-free mul…
- 🪴 prompt maturity 44/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`19c32d6f` docs(pitch): drop the auditable-workspace kicker from the five-pain slide

prompt — I guess we can remove "This is why every agent needs an auditable workspace." from slide 3 in the pitch commit it
- 🪴 prompt maturity 25/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`31e9c308` docs(pitch): lay the five pains in two rows with larger fonts

prompt — If we make those five items locate in two rows, can we make the font size bigger?
- 🌱 prompt maturity 16/100 nascent — vague / under-specified, weak context
model claude-opus-4-8 · agent claude-code

`8422cade` docs(pitch): trim the five-pain explanations to short phrases

prompt — Let's reduce the text explanations in slide 3.
- 🌱 prompt maturity 16/100 nascent — too short, weak context
model claude-opus-4-8 · agent claude-code

`5c66463d` docs(pitch): merge the two problem slides into one five-pain slide

prompt — I guess we can merge slide 3 and 4, no? We can use the style of slide 3, while using the context of slide 4. For the items, I guess we can split them to 5 items, Unversioned prompts, Lost Context, Unsafe and unmonitored autonomy, multi-agent waste, multi-agent conflict …
- 🪴 prompt maturity 32/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`bf3f0881` docs(pitch): reorder to Already-painful before The-problem, restructure into three pains

prompt — "Already painful. The pain is already here." should be the third slide, and then flow is more natural. Also, I guess we can restructue the pains; 1) agent forgot the context, prompt-version control is important to improve prompts/reproduce results 2) agent might destory environem…
- 🪴 prompt maturity 39/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`fd4c68ba` docs(pitch): make slide 2 cards mirrored bullet lists

prompt — can we make "Prompts and commands go untracked, and an agent can run out of control, destroying the environment and conflicting with other agents." to bullet points, like "- untracked prompts and contexts", "- dangerous tool usage", "- conflicting multi-agents". We can also make …
- 🪴 prompt maturity 36/100 developing — no acceptance criteria, weak context
model claude-opus-4-8 · agent claude-code

`0d114ee3` docs(pitch): mirror the 'with h5i' card to the today risks

prompt — in "with h5i", we can say "sandboxed, fully recorded, conflict-free and multi-agent develpment"
- 🪴 prompt maturity 25/100 developing — weak context, no acceptance criteria
model claude-opus-4-8 · agent claude-code

`30f09683` docs(pitch): sharpen the 'today' card on runaway-agent risk

prompt — maybe, `an agnet can run out of control, destroying environemt and conflicting with other agents"?
- 🌱 prompt maturity 21/100 nascent — vague / under-specified, weak context
model claude-opus-4-8 · agent claude-code

`35dcb139` docs(pitch): frame slide 2 cards around safety

prompt — maybe we can say "An agent run is unsafe and opaque" in Today, and "An agent run is safe and auditable" in with h5i. Also, we can mention agent run might be dangenrous in today's section, like "Prompt and commands are not tracked, agents might run out of control.."
- 🪴 prompt maturity 28/100 developing — no acceptance criteria, vague / under-specified
model claude-opus-4-8 · agent claude-code

_{Generated by h5i · re-run h5i share pr post to refresh.}

Koukyosyumei added 30 commits June 30, 2026 10:36

improve hero-badge

96d3c39

docs(pitch): simplify title slide to the one-line definition

3198341

Drop the three feature cards and the closing punchline on slide 1, and enlarge the lede so the h5i definition is the focal point.

docs(pitch): make the title-slide definition three bullet points

053ebff

docs(pitch): enlarge the title-slide bullet points

6269968

docs(pitch): enlarge title-slide bullets further

5918c46

update

ae81631

docs(blog): stop featuring 'Why AI Agents Need Auditable Workspaces'

e343a9c

Demote it from the highlighted featured card to a regular post card in the Ensemble group.

docs(guides): feature the 'Run many agents, merge one auditable resul…

cb7bf2f

…t' overview Apply the shared blog.css featured-card treatment to the workflows overview entry at the top of the guides list.

docs(pitch): add a social-proof line to the title slide

59bfe28

'430+ GitHub stars · the #1 worktree manager by stars' under the definition bullets.

docs(pitch): tighten slide 2 to the two contrast cards

8929718

Remove the lede and the closing summary paragraph. Shorten and enlarge the today/with-h5i cards, and broaden the h5i side from 'evidence' to sandbox + record + multi-agent review (one verified result).

docs(pitch): frame slide 2 cards around safety

35dcb13

today: 'An agent run is unsafe and opaque' (untracked, can run out of control); with h5i: 'An agent run is safe and auditable'.

docs(pitch): sharpen the 'today' card on runaway-agent risk

30f0968

Note that an out-of-control agent can destroy the environment and conflict with other agents.

docs(pitch): mirror the 'with h5i' card to the today risks

0d114ee

Sandboxed, fully recorded, conflict-free multi-agent development.

docs(pitch): make slide 2 cards mirrored bullet lists

fd4c68b

today (✕ red): untracked prompts/context, dangerous tool usage, conflicting multi-agents. with h5i (✓ green): sandboxed environment, tracked prompt/context, conflict-free multi-agent development.

update

bd6a6a0

docs(pitch): trim the five-pain explanations to short phrases

8422cad

update

ea0f192

docs(pitch): lay the five pains in two rows with larger fonts

31e9c30

Switch the pain grid to a centered flex-wrap (3 + 2 rows) and enlarge the item titles (1.9rem) and descriptions (1.2rem).

docs(pitch): drop the auditable-workspace kicker from the five-pain s…

19c32d6

…lide

docs(pitch): recast the product slide as five answers to the five pains

6dce217

docs(pitch): split team detail into prompts/context, env sandbox, and…

059ef69

… team slides

update

3615c8f

docs(pitch): add token-reduction slide with raw-to-filtered comparison

e606da8

docs(pitch): move traction to slide 9 and refresh its metrics

7a4460c

docs(pitch): fix sandbox embed overflow in PDF export

04c7461

update

a0f79e1

docs(pitch): add audit pain/answer pair, an audit-review slide, and r…

1540e52

…efine buyer slide

docs(pitch): reframe closing slides around the AI-code acceptance layer

889a3e0

Koukyosyumei added 5 commits June 30, 2026 23:49

update

56e1fb6

update

848b76e

docs(pitch): restructure beachhead slide into beachhead + expansion t…

d0de02c

…iers

docs(readme): add a short intro line and fix a typo

2bb0197

update

419285b

Koukyosyumei changed the title ~~Improve pitch~~ docs(pitch): refactor the pitch slides Jul 1, 2026

Koukyosyumei merged commit 0b276e0 into main Jul 1, 2026
5 checks passed

Koukyosyumei deleted the improve-pitch branch July 1, 2026 04:05

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

docs(pitch): refactor the pitch slides#245

docs(pitch): refactor the pitch slides#245
Koukyosyumei merged 35 commits into
mainfrom
improve-pitch

Koukyosyumei commented Jul 1, 2026

Uh oh!

Koukyosyumei commented Jul 1, 2026

✅ Security scan clean

`2bb0197a` docs(readme): add a short intro line and fix a typo

`d0de02c4` docs(pitch): restructure beachhead slide into beachhead + expansion tiers

`889a3e0a` docs(pitch): reframe closing slides around the AI-code acceptance layer

`1540e529` docs(pitch): add audit pain/answer pair, an audit-review slide, and refine buyer slide

`04c7461b` docs(pitch): fix sandbox embed overflow in PDF export

`7a4460c7` docs(pitch): move traction to slide 9 and refresh its metrics

`e606da87` docs(pitch): add token-reduction slide with raw-to-filtered comparison

`059ef698` docs(pitch): split team detail into prompts/context, env sandbox, and team slides

`6dce2176` docs(pitch): recast the product slide as five answers to the five pains

`19c32d6f` docs(pitch): drop the auditable-workspace kicker from the five-pain slide

`31e9c308` docs(pitch): lay the five pains in two rows with larger fonts

`8422cade` docs(pitch): trim the five-pain explanations to short phrases

`5c66463d` docs(pitch): merge the two problem slides into one five-pain slide

`bf3f0881` docs(pitch): reorder to Already-painful before The-problem, restructure into three pains

`fd4c68ba` docs(pitch): make slide 2 cards mirrored bullet lists

`0d114ee3` docs(pitch): mirror the 'with h5i' card to the today risks

`30f09683` docs(pitch): sharpen the 'today' card on runaway-agent risk

`35dcb139` docs(pitch): frame slide 2 cards around safety

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

Koukyosyumei commented Jul 1, 2026

Uh oh!

Koukyosyumei commented Jul 1, 2026

🪙 72% AI-authored · 2 files · prompt 44/100

💬 The ask

📍 What shipped

✅ Security scan clean

2bb0197a docs(readme): add a short intro line and fix a typo

d0de02c4 docs(pitch): restructure beachhead slide into beachhead + expansion tiers

889a3e0a docs(pitch): reframe closing slides around the AI-code acceptance layer

1540e529 docs(pitch): add audit pain/answer pair, an audit-review slide, and refine buyer slide

04c7461b docs(pitch): fix sandbox embed overflow in PDF export

7a4460c7 docs(pitch): move traction to slide 9 and refresh its metrics

e606da87 docs(pitch): add token-reduction slide with raw-to-filtered comparison

059ef698 docs(pitch): split team detail into prompts/context, env sandbox, and team slides

6dce2176 docs(pitch): recast the product slide as five answers to the five pains

19c32d6f docs(pitch): drop the auditable-workspace kicker from the five-pain slide

31e9c308 docs(pitch): lay the five pains in two rows with larger fonts

8422cade docs(pitch): trim the five-pain explanations to short phrases

5c66463d docs(pitch): merge the two problem slides into one five-pain slide

bf3f0881 docs(pitch): reorder to Already-painful before The-problem, restructure into three pains

fd4c68ba docs(pitch): make slide 2 cards mirrored bullet lists

0d114ee3 docs(pitch): mirror the 'with h5i' card to the today risks

30f09683 docs(pitch): sharpen the 'today' card on runaway-agent risk

35dcb139 docs(pitch): frame slide 2 cards around safety

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

`2bb0197a` docs(readme): add a short intro line and fix a typo

`d0de02c4` docs(pitch): restructure beachhead slide into beachhead + expansion tiers

`889a3e0a` docs(pitch): reframe closing slides around the AI-code acceptance layer

`1540e529` docs(pitch): add audit pain/answer pair, an audit-review slide, and refine buyer slide

`04c7461b` docs(pitch): fix sandbox embed overflow in PDF export

`7a4460c7` docs(pitch): move traction to slide 9 and refresh its metrics

`e606da87` docs(pitch): add token-reduction slide with raw-to-filtered comparison

`059ef698` docs(pitch): split team detail into prompts/context, env sandbox, and team slides

`6dce2176` docs(pitch): recast the product slide as five answers to the five pains

`19c32d6f` docs(pitch): drop the auditable-workspace kicker from the five-pain slide

`31e9c308` docs(pitch): lay the five pains in two rows with larger fonts

`8422cade` docs(pitch): trim the five-pain explanations to short phrases

`5c66463d` docs(pitch): merge the two problem slides into one five-pain slide

`bf3f0881` docs(pitch): reorder to Already-painful before The-problem, restructure into three pains

`fd4c68ba` docs(pitch): make slide 2 cards mirrored bullet lists

`0d114ee3` docs(pitch): mirror the 'with h5i' card to the today risks

`30f09683` docs(pitch): sharpen the 'today' card on runaway-agent risk

`35dcb139` docs(pitch): frame slide 2 cards around safety