Skip to content

Commit 1b83f43

Browse files
author
aegntic
committed
feat(showcase): interactive harvest knowledge brief - 26 video distillation
1 parent add2e7d commit 1b83f43

2 files changed

Lines changed: 753 additions & 0 deletions

File tree

‎00_Showcase/agent-brief/distilled-index.md‎

Lines changed: 195 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,195 @@
1+
# Harvest Knowledge Brief: 26 Videos, Distilled
2+
3+
Source: 26 YouTube videos harvested and analyzed 2026-09-01. Every claim below traces to a video timestamp. Revenue and performance claims from creators are marked [unverified] - we did not audit them.
4+
5+
**Totals:** 26 videos | 9,112,381 combined views | 1,208 watch-minutes (20.1 hours) | 7 themes | 25 steal-list tactics
6+
7+
**Thesis:** The 2026 playbook across every niche is the same move: point a persistent agent harness at one boring, high-ticket business problem, feed it a distilled knowledge base, sell the outcome on a retainer, and let compounding assets (skills, wikis, microsites, brand files) do the leverage.
8+
9+
---
10+
11+
## Theme 1: Agent Monetization Playbooks (5 videos)
12+
13+
Videos: v01 Higgsfield AI, v02 BRAD LEA TV, v07 Zack Kirk, v13 The Next New Thing, v16 Ben Hawksworth
14+
15+
**Throughline:** Sell agents to boring high-ticket local businesses (roofers, paving franchises, mechanics, dental). Never sell "AI"; sell the outcome (quotes in 20 seconds, more jobs approved). Price structure repeats across channels: roughly 1,000 to 20,000 up front plus 300 to 2,000 per month retainer.
16+
17+
**Sharpest insights:**
18+
19+
1. Quoting agents outsell chatbots. A VIN-decoding tire-and-brake quoter sold for 10-20K up front plus 1-2K/month, because it kills a 30-minute manual pricing chore every call (v07 31:25).
20+
2. Distribution beats building: the partner model routes through marketing agencies who already own the client relationship, at 20% commission (v07 42:38, 58:52).
21+
3. Visual AI mockups crack cold outreach: a before/after photo mockup of the prospect's own parking lot pulled a 3% reply rate and 150K+ of quotes in two weeks for a 25-year paving franchise now up 48% jobs YoY [unverified] (v13 1:13, 2:33, 3:25).
22+
23+
**Contradiction:** "full business in 24 hours" framing (v01) versus the same channel admitting 99% of agent builds is iterative testing (v07 28:22).
24+
25+
---
26+
27+
## Theme 2: Harness Ops: Claude Code as an OS (4 videos)
28+
29+
Videos: v09 Sharbel A., v12 Nate Herk, v18 AI Jason, v24 Nate Herk (2.5h course)
30+
31+
**Throughline:** The operators winning on cost are not prompting better; they are managing context state: clearing, deferring tool manuals, distilling wikis, and scheduling around cache windows.
32+
33+
**Sharpest insights:**
34+
35+
1. Spend is rereads, not requests: one maker measured 96% of spend as re-sending conversation history; /clear between jobs resets the base to zero and costs nothing (v09 1:30, 3:02, 3:49).
36+
2. Sub-agents do not save tokens, they move them; Anthropic's own docs show the win is isolated clean contexts, not volume discounts (v09 10:15, v12 21:47).
37+
3. Distill raw sources into a wiki: transcripts compressed into a compact wiki dropped token usage on queries by 95% [unverified] (v24 131:19).
38+
39+
**Contradiction:** micro-optimizing every token (v09) versus running 2.5-hour daily multi-agent content engines on the same subscriptions (v23, v24). Both are real strategies; they optimize for different constraints.
40+
41+
---
42+
43+
## Theme 3: Hermes and Open Harnesses (3 videos)
44+
45+
Videos: v04 Lead Gen Jay, v14 The AI Architects, v20 zSecurity
46+
47+
**Throughline:** Harness choice beats model choice on cost and completion. A VPS-hosted Hermes reached via Telegram/iMessage, wired to a cheap subscription, outperforms per-seat vendor agents for solo operators.
48+
49+
**Sharpest insights:**
50+
51+
1. In head-to-head use, a 300/month Grok Bot subscription lost to Hermes as the better harness, while Hermes ran on a 20/month ChatGPT subscription (v04 9:40, 10:27).
52+
2. Composio's harness benchmarks (highest completion rate, cheapest cost per task, fastest runtime vs Claude Code) [unverified] are the argument for treating the harness as the primary stack decision (v14 2:56, 3:15, 3:50).
53+
3. One phone message can fan out a parallel hacking team (SMS spoof, tracking page, dashboard) with sub-agents that notify you on Telegram when blocked (v20 0:00, 23:42).
54+
55+
**Contradiction:** none internal; the tension is vendor convenience (Grok Bot, Hyper Agent) versus open harness economics (Hermes). v04 pays 300/month and still concludes Hermes wins.
56+
57+
---
58+
59+
## Theme 4: Traffic, SEO and Microsites (4 videos)
60+
61+
Videos: v08 Jonathan's Jam, v10 The Next New Thing, v21 Jesse Cunningham, v22 Jack Roberts
62+
63+
**Throughline:** Compounding utility assets beat the content treadmill: tool sites and microsites built once keep earning, and agents collapsed their build cost to near zero.
64+
65+
**Sharpest insights:**
66+
67+
1. A word-unscrambler site doing 55M monthly visits at an assumed 9 RPM implies ~495K/month [unverified]; the durable insight is that utility demand never expires (v08 2:40, 3:01, 3:46).
68+
2. The takeover formula: one industry x 5 sub-niches x 20 metro cities = 100 exact-match-domain flat-HTML microsites, each with a unique phone number to attribute winners, rented per lead (v21 4:44, 5:35, 10:05).
69+
3. Agents doing SEO without real data just guess; wiring an open-source SEO data MCP (10/month hosted, or 50 of DataForSEO credits that can last months) grounds them (v10 1:44, 2:03, 3:14).
70+
71+
**Contradiction:** "1,000 microsites" bravado (v21) versus the same channel's own rule: prove one, sell it, then scale. Volume is the marketing; sequence is the method.
72+
73+
---
74+
75+
## Theme 5: Content Machine and Personal Brand (4 videos)
76+
77+
Videos: v03 DevJesse, v11 Diary Of A CEO, v19 Calum Johnson Show, v23 Grow with Alex
78+
79+
**Throughline:** Winning creators do not ask AI for ideas; they install an engine (brand files, winner analysis, reference frames) and keep human judgment on the final 10%.
80+
81+
**Sharpest insights:**
82+
83+
1. Only ~1% of platform users create; the competition is far thinner than it looks, and posting compounds everything else (v11 12:46, 28:17).
84+
2. Study winners at the thumbnail/title level, not the full video: 90% of a YouTube video's fate is cover plus title, so scrape those across the niche instead of watching everything (v19 54:04).
85+
3. Brand files feed everything: five gathered docs (audience, voice, visual, offer, existing brand) turn one-off prompting into a repeatable engine; recurring jobs become skills (v23 31:01, 82:34).
86+
87+
**Contradiction:** "AI content is slop" (v19 8:41) versus entire channels cloned with a free Claude plan (v03). Resolution from v19 itself: AI drafts 90%, a creative director finishes the last 10.
88+
89+
---
90+
91+
## Theme 6: GTM, Apps and the Marketing Engineer (2 videos)
92+
93+
Videos: v15 Greg Isenberg, v25 Viral App Launch
94+
95+
**Throughline:** Growth now means wiring agents directly into live business data (sales calls, funnels) and instrumenting monetization before chasing users.
96+
97+
**Sharpest insights:**
98+
99+
1. Monetize before you scale: 46,000 downloads with zero monetization collapsed a startup; 5-10% conversion would have been ~46K/month (v25 2:42, 2:57).
100+
2. The marketing engineer role (one person + agent systems doing a marketing team's work) is priced at 250K to 1M/year because it sits next to revenue (v15 0:17, 16:45, 25:51).
101+
3. A growth repo: markdown files that update from sales calls each week and generate founder posts, videos, and landing pages from real objections (v15 8:16, 18:53, 21:03).
102+
103+
**Contradiction:** "stop building, freeze the product" (v25 4:01) versus design-skill channels shipping polished sites weekly (v22). Different businesses; the rule binds app founders, not service studios.
104+
105+
---
106+
107+
## Theme 7: Trading-Adjacent Discipline (2 videos)
108+
109+
Videos: v06 IQCapital, v17 DaviddTech
110+
111+
**Throughline:** The two trading videos are really about process discipline, and they map cleanly onto agent ops: journal everything, forward-test before going live, and size risk so one failure cannot kill you.
112+
113+
**Sharpest insights:**
114+
115+
1. The Robbins Cup champion's edge is trapped-participant logic (failed moves, not predictions) plus collecting his own trade data to find time-of-day weaknesses (v06 6:44, 46:10).
116+
2. A 22-hour Grok trading bot run on 10K produced 6 trades, 3 wins, 3 losses, +11.60 total (v17 17:04); the honest lesson is the maker's own: 99% of backtests fall apart, so journal and forward-test for 3 months before trusting any automated system (v17 10:36, 19:50).
117+
3. Challenge-account math (2K drawdown vs 3K target) forces a 1.5R-at-65%-win-rate style, not hero risk (v06 33:20, 33:42).
118+
119+
**Contradiction:** a 100% monthly championship return [unverified] (v06 2:50) versus the agent reality of +11.60 in a day (v17). Both are honest videos; the gap between champion discretionary skill and autonomous bots is the finding.
120+
121+
---
122+
123+
## Cross-Theme Contradictions
124+
125+
1. Passive income framing ("runs without me", v05; "$1M/yr run entirely by AI", v02) versus operator testimony that agents need constant testing and updating (v07 28:22, 54:42) and that 67% of employees admit sabotaging AI rollouts [unverified] (v02 50:56).
126+
2. Volume plays (hundreds of ad variants in v01 7:38; 1,000 microsites in v21) versus focus doctrine (prove one first in v21 11:26; "one offer, one niche" in v01 0:53).
127+
3. Cheap-stack ideology (Hermes on 20/month, v04) versus the same operators paying 300/month for Grok Bot just to benchmark it (v04 0:00).
128+
4. RPM-based utility-site income math (v08) is presented as near-passive, but the same video concedes first-page ranking is "a whole industry" (v08 12:27) [unverified].
129+
130+
---
131+
132+
## Steal List: 25 Tactics, Ranked by Leverage x Cheapness
133+
134+
Ranked for a solo agent studio: highest leverage and lowest cost first. Each row names the source videos and the one-line how-to.
135+
136+
| # | Tactic | How (one line) | Leverage | Cheap | Sources |
137+
| --- | ------------------------------ | ------------------------------------------------------------------------------------------------------------ | -------- | ----- | --------------------- |
138+
| 1 | Context reset ritual | /clear between jobs; audit what resends on every turn before optimizing anything else | 5 | 5 | v09 3:02 |
139+
| 2 | Wiki distillation loop | Nightly job compresses raw transcripts and notes into a wiki; query the wiki, never the raw pile | 5 | 5 | v14 15:19, v24 131:19 |
140+
| 3 | Brand-file engine | Maintain audience/voice/visual/offer docs and require every generation to read them first | 5 | 5 | v15 8:16, v23 31:01 |
141+
| 4 | Skills for recurrence | When a prompt runs twice, freeze it into a skill with docs; prompts are for one-offs | 4 | 5 | v23 82:34, v24 93:28 |
142+
| 5 | Cheap harness stack | Run an open harness on a 20/month subscription instead of 300/month vendor seats | 5 | 5 | v04 10:27, v14 3:15 |
143+
| 6 | Boring-niche selector | Ask the model: which niches spend big on ads but have the worst digital quality; skip clothing and coffee | 4 | 5 | v01 0:53 |
144+
| 7 | Outcome offer formula | Offer = result + real problem + risk killed + speed + bonus + explicit next action | 4 | 5 | v01 3:51 |
145+
| 8 | Winner-signal scraping | Scrape thumbnails and titles of top performers in the niche instead of watching full videos | 4 | 5 | v19 54:04, v03 8:36 |
146+
| 9 | Growth repo from calls | Pipeline sales-call transcripts into markdown insights that generate posts, videos, and landing pages weekly | 5 | 4 | v15 18:53, v19 43:57 |
147+
| 10 | Council red-team | Spin up adversarial personas (contrarian, expansionist, first customer) to kill weak ideas before building | 4 | 5 | v12 3:07 |
148+
| 11 | Agent self-verification | Make the agent screenshot its own UI states and probe edge cases before you look | 4 | 5 | v12 9:53, v07 28:22 |
149+
| 12 | Sub-agent clean contexts | Fan out sub-tasks to isolated contexts and end sessions immediately after they report | 4 | 4 | v12 21:47, v09 10:15 |
150+
| 13 | Microsite takeover grid | 1 industry x 5 sub-niches x 20 cities; flat HTML, exact-match domains, unique phone number per site | 5 | 3 | v21 4:44, 5:35 |
151+
| 14 | Utility tool compounding | Ship small calculator/generator tools on easy keywords; AdSense them and let them sit | 4 | 4 | v08 13:02, 18:45 |
152+
| 15 | Monetization first | Instrument funnel steps and turn on payments before spending a dollar on acquisition | 5 | 5 | v25 5:20, 6:18 |
153+
| 16 | Trigger-based cold email | One line referencing a real trigger event (funding, new hire, pothole photo mockup); no pitch | 4 | 4 | v16 5:39, v13 1:13 |
154+
| 17 | Retainer-priced agents | Sell the build at 1-20K plus 300-2,000/month for updates, data feeds, and accuracy | 5 | 4 | v07 54:42, v16 3:28 |
155+
| 18 | Partner distribution | Route sales through agencies that already own the client; pay 20% commission | 5 | 4 | v07 42:38, 58:52 |
156+
| 19 | Grounded SEO data | Wire an SEO data API via MCP so the agent stops hallucinating keywords and backlinks | 4 | 4 | v10 1:44, 3:14 |
157+
| 20 | Channel-clone system | Feed 3 top videos' titles and transcripts to derive a style guide, name, description, and format you repeat | 3 | 5 | v03 4:47, 12:45 |
158+
| 21 | Template referral assets | Publish sites as remixable templates; earn referral fees when others fork and subscribe | 3 | 4 | v08 14:16 |
159+
| 22 | Scheduled-cache hygiene | Run cron tasks at 45min+ intervals; cache misses reprice the whole context at 10x | 3 | 5 | v09 13:42, 17:06 |
160+
| 23 | Forward-test journal | Before trusting any automated system, paper-trade or journal it for 3 months; expect backtests to lie | 4 | 5 | v17 10:36, 19:50 |
161+
| 24 | Multi-agent parallel sessions | Run 10-15 isolated agent sessions via a control graph with deterministic checks between nodes | 4 | 3 | v18 3:36, 18:06 |
162+
| 25 | AI-draft human-finish pipeline | Let agents draft 90%; reserve human judgment for the final 10% where taste lives | 4 | 4 | v19 11:01 |
163+
164+
---
165+
166+
## Video Index
167+
168+
| # | Title | Channel | Views | Length | ID |
169+
| --- | --------------------------------------------------------------------------------- | -------------------------------- | --------- | ------ | ----------- |
170+
| 00 | The AI Agent Every Company is About to Build (Vercel CEO) | Riley Brown | 94,171 | 54:22 | HQXi4snP36I |
171+
| 01 | Zero to $13,457/Mo: 24-Hour AI Agency Experiment [unverified] | Higgsfield AI | 101,108 | 23:34 | iOwKylW8c5Q |
172+
| 02 | The Hidden $1M/Year Business Model Run Entirely By AI [unverified] | BRAD LEA TV | 31,585 | 82:10 | dIiRTuLKImk |
173+
| 03 | I Cloned a $10,000/Month YouTube Channel Using Claude Code For Free [unverified] | DevJesse | 29,700 | 25:11 | IS0mJP_7dZE |
174+
| 04 | Everything You Need to Know About Grok Bot (Full Grok Build) | Lead Gen Jay | 21,050 | 17:45 | v8xFG4RmsLU |
175+
| 05 | How to literally print money while you sleep with AI dropshipping [unverified] | Romas Ecom | 154,391 | 25:59 | ScF0pi6FSLQ |
176+
| 06 | Trading WORLD CHAMPION Reveals the Orderflow Strategy (Robbins Cup) | IQCapital | 874,681 | 58:00 | PL7LKUsCgIQ |
177+
| 07 | This ONE AI Agent Makes me $96,000/Month [unverified] | Zack Kirk | 40,445 | 60:38 | L1lswXTpLjw |
178+
| 08 | This Boring Website Earns $6M/Year. How to Clone it [unverified] | Jonathan's Jam | 750,503 | 20:27 | nQOGK72IHx8 |
179+
| 09 | Paste This Into Claude Code, Never Run Out Of Tokens Again | Sharbel A. | 136,718 | 19:47 | kHtOSJRUkLs |
180+
| 10 | New: your agent lands customers (open-source SEO) | The Next New Thing | 50,087 | 19:02 | UNN6dLt5dPI |
181+
| 11 | The 7,11,4 Hack That Turns $1 Into $10K Per Month (Daniel Priestley) [unverified] | The Diary Of A CEO | 5,680,756 | 132:32 | sFkR34AMPw8 |
182+
| 12 | I asked Claude Code to make me as much money as possible | Nate Herk / AI Automation | 201,316 | 28:12 | iTY8Q449YNQ |
183+
| 13 | A little-known agent is shockingly powerful (paving franchise) | The Next New Thing | 8,214 | 23:01 | Tj7uyTCX4-4 |
184+
| 14 | How To Build The ULTIMATE AI Second Brain for Hermes Agent | The AI Architects / Tom Crawshaw | 30,750 | 26:10 | wvYAuHfJRo0 |
185+
| 15 | Marketing Engineer: The $1M Job with AI Agents [unverified] | Greg Isenberg | 31,570 | 35:18 | 8ZC1G1ezN5o |
186+
| 16 | 7 AI Businesses You Can Start This Weekend (Just Claude) | Ben Hawksworth | 22,994 | 15:15 | wudOmpK-8Fk |
187+
| 17 | I Gave the Grok Bot AI Trading Bot $10,000 for 22 Hours | Trading with DaviddTech | 42,112 | 21:23 | muHeyXiCn-A |
188+
| 18 | I don't prompt agents anymore (graph engineer) | AI Jason | 26,186 | 22:22 | _9OT25ZvrWs |
189+
| 19 | The AI INFLUENCER EXPLAINS: AI Agents for Content & Brand [unverified] | The Calum Johnson Show | 38,262 | 98:45 | DJw2yeQAPXc |
190+
| 20 | I Built an AI Hacking Team with Hermes Agent | zSecurity | 206,495 | 29:11 | zwV5p1L0COI |
191+
| 21 | How to Build 1,000 AI Microsites and Dominate an Entire Industry | Jesse Cunningham | 63,480 | 14:25 | moS5SjJFkt4 |
192+
| 22 | Insane Claude Design Skills You Need for $10,000 Websites [unverified] | Jack Roberts | 10,900 | 22:56 | pUu4G2lINnk |
193+
| 23 | Claude For Social Media: The FULL Course (2.5 Hours, Free) | Grow with Alex | 14,457 | 157:12 | hoVC2W0p0Zg |
194+
| 24 | Build & Sell Claude Code Operating Systems (2+ Hour Course) | Nate Herk / AI Automation | 426,539 | 153:21 | bCljOfCH8Ms |
195+
| 25 | How To Get 10,000 App Users In 6 Months (Full System) | Viral App Launch | 23,911 | 20:43 | cgmBuqkkUQc |

0 commit comments

Comments
 (0)