google search grew 12%. google cloud grew 103%.
people didn't stop searching. they started asking.
if your GTM still lives on blue links, here's what i'd build this quarter:
1. answer pages: one page per question your buyer asks chatgpt. plain answer in the first 2 lines.
2. comparison pages: "you vs X" for every competitor. models love clean tables.
3. llms.txt + docs that load without JS so agents can actually read you.
4. third party mentions: reddit, G2, podcasts, newsletters. models cite who others cite.
5. a friday check: run 20 buyer prompts in chatgpt, claude and perplexity. log who gets named. that's your new rank tracker.
SEO isn't dead. the click is.
1
74
The organic-first loop that finds winners before you spend on ads:
1. Make 10 content variations (AI for scripts, hooks and UGC style video)
2. Post them organically across TikTok, Reels and Shorts, same angle, different hooks
3. Keep posting daily for a full month, that's how long it takes to see real patterns
4. Flag anything that consistently beats your usual views, saves or signups
5. Take the survivors and make 5 more versions of each, same format, new hook
6. Post those too and keep the formats that win again
7. Only then put paid budget behind the proven videos, at 2 to 3x
Most people pay for ads on untested creative.
Organic is the free test lab. Paid is just the megaphone for what already works.
3
3
144
Props to @shnufalo_ for making @revivosupps
Have been having the mango electrolytes and damn this is sooo good!!!
Never thought I would love any electrolytes, but these are 🤌🏻🤌🏻
Go check it out🥭🥭
Had to rotate every single key I have on Railway this week.
Railway got breached. Up to ~100,000 workspaces had env vars potentially exposed, and the attacker was hunting AI API keys.
Same week:
PHP shipped emergency patches for two critical bugs that let attackers run code on a server with no login.
Node.js patched 10 CVEs.
One hacker pointed AI agents at 1,200+ sites and got into 100+ companies for about $50 per hack.
Everything is on fire right now. If you're a founder, steal this before the next one:
hack-week/
├── first-hour/
│ ├── list every key: AI, Stripe, DB, GitHub, email
│ ├── rotate AI keys first, that's what they're hunting
│ ├── regenerate DB passwords, turn off public DB access
│ └── redeploy so the new keys actually load
├── secrets/
│ ├── one key per app, never one key for everything
│ ├── spend cap on every AI key
│ ├── reference variables, not pasted values
│ └── no keys in Slack, Notion or shared .env files
├── dependencies/
│ ├── pin exact versions, never "latest"
│ ├── patch Node and PHP the day fixes ship
│ └── new package versions wait a few days before prod
├── customers/
│ ├── tell them before they ask
│ ├── what happened, what you did, what they should do
│ └── one email, plain words, no legal speak
└── plan-b/
├── a second place you can deploy in a day
└── a written list of who rotates what
The breach isn't what customers remember. The silence is.
4
1
80
One dev's first iOS app did $41K last month. No TikTok. No paid social. Apple Search Ads and boring SEO pages.
The loud channels get the likes. The boring ones pay rent.
Steal this boring-channel stack:
boring-growth/
├── app-store/
│ ├── Apple Search Ads on exact-match competitor + problem keywords
│ ├── keyword in title + subtitle, not just description
│ ├── 3 screenshots that sell the outcome, not the UI
│ └── ask for a review right after the first win moment
├── seo/
│ ├── 1 page per "how to [problem]" query
│ ├── comparison pages: you vs every competitor
│ ├── refresh old posts ranking 8 to 20
│ └── every page ends in a download link
├── ai-search/
│ ├── answer the exact questions people ask ChatGPT
│ ├── get listed in every "best X apps" roundup
│ └── check monthly if ChatGPT recommends you
├── lifecycle/
│ ├── day 1, 3, 7 push or email tied to the core action
│ └── win-back at day 30
└── math/
├── CPA per keyword, weekly
├── kill any keyword above 1/3 of LTV
└── double down on the 20% that pay
Boring compounds. Viral doesn't.
1
54
One founder hit about a $25M run rate in under a year. The early growth wasn't magic, it was emails to old users, X replies and ads.
Steal this distribution stack:
distribution/
├── owned/
│ ├── email your old users every launch
│ ├── waitlist with a referral bump
│ └── changelog that reads like a story
├── social/
│ ├── 20 replies a day in your niche
│ ├── 1 build-in-public post daily
│ └── DM everyone who engages twice
├── paid/
│ ├── start at $20/day on one channel
│ ├── kill any ad that misses CAC in 72h
│ └── turn winning replies into ad copy
├── content/
│ ├── 1 teardown a week of a viral post
│ └── repurpose every thread into 3 formats
└── loops/
├── share card in the product
└── ask for the review right after a win
Distribution is a daily habit, not a launch day.
4
41
Steal this agent verification layer.
Everyone copies the "run 20 agents in parallel" setup.
Almost nobody copies the part that makes it work: the layer that refuses to believe them.
Before any agent output touches a customer or the CRM:
1) Receipts or it didn't happen
Every "done" needs a link, a diff, or a log line. No receipt = not done.
2) Replay on old data first
New agent or new prompt runs on last week's leads and tickets before it ever sees today's.
3) A second agent grades the first
Cheap model checks: did it follow the rules, invent a fact, skip a step. Score 1-5. Under 4 goes to a human.
4) Write access is earned
Read-only for 7 days. Then drafts only. Then real writes with a daily cap.
5) One kill switch per agent
Test it on day one, not on the bad day.
6) Friday diff review
15 min: what each agent changed, what I had to fix, which prompt gets rewritten.
2
30
📂 first-10-customers
┣ 📂 icp
┃ ┣ who.md
┃ ┣ not-who.md
┃ ┗ pain.md
┣ 📂 triggers
┃ ┣ hiring
┃ ┣ funding
┃ ┗ “just launched”
┣ 📂 list
┃ ┣ accounts.csv (50)
┃ ┗ disqualified.csv
┣ 📂 messaging
┃ ┣ angle-v1
┃ ┣ angle-v2
┃ ┗ angle-v3
┣ 📂 channels
┃ ┣ cold-email
┃ ┣ linkedin
┃ ┗ warm-intro
┃ ┗ (pick ONE)
┣ 📂 ops
┃ ┣ daily-sends
┃ ┣ followups/
┃ ┗ crm-notes/
┗ 📂 weekly
┣ what-worked.md
┣ what-died.md
┗ next-hypothesis.md
3
48
Your agents are only as smart as last week’s context.
Steal this Weekly Context OS:
MON: Close + aim
• 3 outcomes for the week (“by Friday I will have…”)
• Flag decisions that need a human owner
TUE: Ingest
• Pull last 7 days: calls, drafts, Slack decisions, win/loss notes
• Extract new facts
• Flag contradictions vs existing truth files
WED: Update truth
• Patch company / customer / offer / voice
• Human approves diffs (agents propose, you sign)
THU: Ship with context
• Every agent task must read the updated truth first
• No blank-chat work for client or GTM output
FRI: Health + memo
• What landed vs the 3 outcomes
• One-sentence reason for misses
• Monthly check: contradictions, pages untouched 60+ days, missing topics
Rules:
• Stale context > no context (because agents trust it)
• Weekly ingest compounds; one-time dump dies
• Context beats another plugin
Set the Friday memo before you buy another AI tool.
2
40
The whole one-person company playbook on one page:
1. Pick one painfully narrow problem
2. Talk to 10 real buyers (not Twitter opinions)
3. Write the case in their language
4. Ship one clear page
5. Close the first customers manually
6. Automate only what already repeats
AI can own:
research · content drafts · sales prep · support drafts · ops gruntwork
You still own:
taste · pricing · “yes/no” · relationships
That’s the split.
Judgment stays human. Repetition gets agents.
Save this. Start with step 1 today.
2
40
📂 agent-outbound
┣ 📂 01_signals
┃ ┣ hiring
┃ ┣ funding
┃ ┣ product-launch
┃ ┣ public-pain
┃ ┗ do-not-chase.md
┣ 📂 02_qualify
┃ ┣ icp-fit.score
┃ ┣ live-reason.md
┃ ┣ budget-signal
┃ ┗ reject-rules.md
┣ 📂 03_enrich
┃ ┣ email
┃ ┣ linkedin
┃ ┣ company-context
┃ ┗ sources.md
┣ 📂 04_draft
┃ ┣ hypothesis.md
┃ ┣ variants/
┃ ┣ proof-lines/ (approved only)
┃ ┗ banned-phrases.md
┣ 📂 05_send
┃ ┣ approval-queue/
┃ ┣ daily-cap.md
┃ ┣ sequences/
┃ ┗ no-auto-send
┗ 📂 06_learn
┣ replies/won
┣ replies/lost
┣ hooks-rank.md
┗ kill-list.md
1
27
Don’t hire 13 SDRs.
Hire 13 roles your agents can run:
STRATEGY
• Head of Sales — WHO + WHY this week
PROSPECTING
• Signal Hunter — live buying intent
• ICP Analyst — kill bad fits early
• Intent Scorer — rank “likely to buy now”
INTEL
• Account Researcher — why them, why now
• Enricher — real email/phone only (no inventing)
OUTREACH
• Copywriter — message from the signal
• Outreach Op — send on human approval
CONVERSION
• Reply Agent — detect interest
• Follow-up Agent — keep warm alive
• Meeting Qualifier — interest → demo-ready
OPTIMIZATION
• Pipeline Analyst — what actually converts
• Sales Manager — coordinate the loop
The kill list that saves your week:
✗ no ICP match → drop
✗ ICP match, zero intent → nurture only
✗ intent but unreachable → enrich first
✗ reachable + intent → outreach now
Flow:
signal → ICP gate → research → enrich → score → message → send → reply → qualify → book → repeat
Objective in. Calendar out.
1
31
Your agent skills fail quietly.
Steal this 9-step hardening checklist:
1) Diagnose → Fix
Paste failing prompt + failure notes.
Map each failure to a structural change.
Show before/after. Re-score on the same 5 tests.
2) Design for failure first
Name the 3 most likely breakage modes.
Preload constraints, format locks, guardrails.
Stress with 5 edge inputs before you ship.
3) Adversarial QA
Generate 10 realistic break attempts.
Log where it degrades.
Severity-rank before you edit anything.
4) Kill vague rules
“Be concise” → “≤100 words”
Every instruction must be testable.
5) Lock the output shape
Named sections. Fixed order. Numeric length rules.
Run 5 times — structure must match.
6) One prompt → multi-phase system
Input → process → validate → output.
Explicit handoffs + checkpoints between phases.
7) Pre-ship reliability score
Clarity · format · constraints · edge cases · tone
Anything <7 = launch risk.
8) Cut weakening words
Flag “try to / if possible / maybe”
Shorter only if clearer.
9) Add a feedback loop
Log failures → turn into active constraints → retest.
Ship skills like production code — or they’ll act like demos.
22
Stop re-explaining your context every chat.
Set these up once:
/skills (teach once → invoke forever)
/who-i-am → your bio, offer, constraints
/no-ai-slop → banned words + patterns
/build-skill → the meta-skill that writes your other skills
/projects
One project per client or workstream.
Upload past work once → every chat inherits it.
Share the project → team stops re-briefing from zero.
/interview-first
Tell the model to ask clarifying questions until gaps are gone.
Click answers > guesswork prompts.
Incomplete inputs are why outputs fail.
/connectors
Wire the tools you actually use (mail, Slack, notes, docs).
Pull live context instead of pasting it.
Bookmark this.
2026 winners run systems. Everyone else retypes the same brief.
23
Steal this: the 4 files every AI workspace needs
company.md
customer.md
offer.md
voice.md
Most people never define what goes inside.
1) company.md
What you sell, who you serve, how money comes in, what you believe, what makes you different.
Fast build: website + last 3 proposals + 1 call transcript → draft in Claude → you fix the 20%.
2) customer.md
ICP, pains, objections, triggers, questions, exact language, fears, decision criteria.
Must come from transcripts — not your imagination.
No quotes = fiction.
3) offer.md
Packages, deliverables, pricing logic, proof, promises, claims to avoid, good-fit rules.
You write this first. Then have AI stress-test ambiguity + skeptical buyer questions.
“Claims to avoid” stops agents from overpromising.
4) voice.md
How you sound / never sound, phrases in / phrases out, good writing samples.
Feed 10 posts you love + 3 you hate → extract rules.
Adjectives (“confident, friendly”) are useless. Examples + bans win.
Put all 4 in one folder. Point every agent at it.
Refresh weekly from new calls + drafts.
Stale context is worse than empty context.
1
20
Every GTM agent setup fails without a shared brain.
Here’s a clean marketing knowledge OS you can clone:
1) Shared truth files (one source for every agent)
> company.md — what you sell + who it’s for
> customer.md — pains, triggers, objections, THEIR words
> offer.md — packages, pricing logic, claims you can defend
> positioning.md — why pick you vs alternatives
> voice.md — examples + banned phrases (not adjectives)
> proof.md — approved claims + evidence links
Build from: calls, proposals, site copy, past campaigns.
2) One folder per function (content / SEO / paid / email / creators)
Inside each:
> playbook.md
> examples/ (approved + rejected + why)
> tools.md
> checks.md (pre-ship QA)
> calendar.md
3) Split content by channel
content/
├── playbook.md
├── checks.md
├── x/
├── linkedin/
├── newsletter/
└── visual/
4) Campaigns get their own folder
campaigns/launch-name/
├── brief.md
├── research/
├── decisions.md
├── production/
└── results.md
5) AGENTS.md = operating rules
where to look · what to trust · where drafts go · what needs human sign-off
6) Compound the brain
Fix a draft → update examples.
New objection → update customer.md.
Offer change → update offer.md + flag live campaigns.
Start with ONE function you’re already running this week.
1
21
Last event was super cool. Excited for this new meetup and all the cool use cases 🔥🔥
Replying to @bot
39
📂 SaaS content + SEO OS (steal the tree)
┃
┣ 📂 Strategy
┃ ┣ keyword map
┃ ┣ search intent
┃ ┣ competitor gaps
┃ ┣ topic graph
┃ ┗ 90-day roadmap
┃
┣ 📂 On-page
┃ ┣ titles / metas
┃ ┣ headings / URLs
┃ ┣ internal links
┃ ┗ image alt + compression
┃
┣ 📂 Content engine
┃ ┣ pillars + clusters
┃ ┣ evergreen refresh
┃ ┣ programmatic pages
┃ ┗ CTR rewrite loop
┃
┣ 📂 Technical
┃ ┣ crawl / index
┃ ┣ sitemap / robots / canonicals
┃ ┣ schema
┃ ┗ CWV + speed
┃
┣ 📂 Off-page
┃ ┣ linkable assets
┃ ┣ digital PR
┃ ┣ broken-link reclaim
┃ ┗ brand mention chase
┃
┣ 📂 SaaS BOFU
┃ ┣ feature / use-case pages
┃ ┣ integrations
┃ ┣ alternatives + comparisons
┃ ┗ free tools
┃
┣ 📂 AI search
┃ ┣ AI Overviews readiness
┃ ┣ citation / entity coverage
┃ ┗ ChatGPT / Perplexity presence
┃
┣ 📂 Measurement
┃ ┣ GSC + GA
┃ ┣ rank + conversion
┃ ┗ weekly experiment log
┃
┗ 📂 Growth
┣ wins backlog
┣ content scale
┗ automation candidates
Save this. Fill one folder per week.
1
1
20



