Abhishek Shankar's Blog
Posts
Ideas
The Map
About
Subscribe
↗
×
Posts
Ideas
The Map
About
Subscribe
↗
Ideas
35 frameworks. Each is the smallest unit of reasoning I'd hand to someone else to make a decision.
all · 35
system · 18
framework · 15
strategy · 11
ai · 11
product · 9
leadership · 8
decision-making · 8
investing · 6
marketing · 5
mental-model · 5
venture · 4
brand · 4
work · 3
knowledge · 3
productivity · 3
delegation · 3
content · 3
distribution · 3
capital · 2
technology · 2
risk · 2
agents · 2
workflow · 2
growth · 2
communication · 1
writing · 1
psychology · 1
innovation · 1
Strategy & Decisions
Risk & Uncertainty
Verification Trust vs Brand Trust
Brand trust is replicable with marketing budget. Verification trust requires years of evidence. The distinction matters most in AI, where the model layer commoditizes weekly.
Positioning & Pricing
Tier-Mapped Sales Motions
SMB, mid-market, and enterprise are not the same business with different price tags. Each tier needs its own qualifying form, CTA, and content shape.
Competitive Strategy
Distribution Is the Meta-Moat
The only durable advantage flows through one channel: who controls the path to the customer. If you'd shut down sales for 90 days, would customers still find you?
Decision-Making
The Three PMF Archetypes
Hair on Fire, Hard Fact, Future Vision. Each archetype requires a different first sale, a different early customer, and a different content engine.
Frameworks & Mental Models
Analytical Frameworks
The Terrifying Questions
Four questions, in order: do people pay? does it create value? can you reach them cheaply? can you defend it? Most founders only ask the first two.
The IDEO Triad
Desirability, feasibility, viability. A product that hits two of three is a project, not a business. Score honestly on the lowest dimension.
Three Horizons Allocation
H1 extends the core, H2 builds adjacencies, H3 bets on breakthroughs. 70/20/10 of attention is the workable allocation. The harder discipline is killing what's underperforming.
Heuristics
The Mom Test Filter
Talk about their life, not your idea. Ask about past behavior, not hypothetical futures. Look for commitments, not compliments. The diagnostic: did they give up something real?
Product & Design
Craft & Quality
Knowledge-First Generation Order
Elicit domain knowledge before any model writes a word. The default flow patches generated text with facts. The knowledge-first flow refuses to write until the facts are in.
Product Sense
Causal vs Lagging PMF Signals
Most PMF definitions are lagging indicators. The causal moment is conversion from cold landing page to first paid action, measured before anyone tells you you have PMF.
Product Discovery
Stated vs Revealed Preference
Stated preference is corrupted signal. Revealed preference is what someone actually did when they thought no one was measuring. If you can't construct one before you build, you're stalling.
Shadow-Mode Validation
Tesla's perception loop applied to startups: stop asking customers what they would do. Run the new behavior in shadow and watch the divergence.
Growth, Brand & Distribution
Brand Building
Hidden Tag Drift
Ship the tag data and the UI surface together, or don't ship the data. The middle state — data exists, UI doesn't — is the worst of both worlds.
Content & SEO
Dual-Optimization Content
Two readers now: Google's algorithm and AI reasoning engines. Front-load keywords for ranking, structure for citation, sources for both. Writing for one audience is over.
The YMYL Authority Tax
Healthcare, finance, and law content won't rank without named credentialed reviewers and EEAT signals. Pay the authority tax upfront, or pay it forever in lost traffic.
Channels
LTV-Driven Channel Choice
LTV picks the channel, not the other way around. If no channel runs profitably for two years, you have an LTV problem — expand the offer until the math works.
Near-Me Keyword Economics
Near-me keywords route already-convinced buyers. Which monetization role wins is decided by two variables: patient LTV and competitive density.
Operations & Systems
Automation
The Causal Knowledge Handoff
Tasks are linear; reasoning compounds. The diagnostic: would this person know what to do if the brief changed by 20%? Hand off the model, not just the actions.
Workflows
Context Sandboxing
Strict memory isolation between domains with explicit bridges. The default is sealed. The difference between a tool you trust and one that surprises you in front of a client.
Risk-Scored Human-in-the-Loop Gates
Score each agent action on reversibility, cost, and scope. Low-risk auto-executes; high-risk pauses with a context summary. Turn the big trust question into many small ones.
Delegation
The 70% Rule
If someone can do it 70% as well, delegate. Anything higher keeps you doing the work forever. Capacity comes from delegation, not before it.
The Eight Levels of Delegation Autonomy
Eight levels from "gather info, I decide" to "they own it completely." Name the level explicitly at handoff, or both sides default to the level that protects them.
The Delegation Gate
A forced pause before any task: can someone else do this, and if so, who? The friction is the point. Without the gate, every task gets done by the most expensive person available — you.
AI & Technology
Architecture Patterns
The Orchestration-Ownership Rule
Use vendor APIs for data, vendor models for inference, but own the orchestration. When the beta gets deprecated, you ship a config change, not a re-architecture.
Hybrid RAG: Dense + Sparse + RRF
Pure vector search misses domain terms. Hybrid — dense + sparse + RRF fusion — is a correctness fix, not an optimization. Dense-only RAG leaves recall on the table.
LLMs & Agents
Documentation as Source-of-Lies
Repo context files often hurt agent success more than help. Docs drift, code moves on, agents follow stale instructions confidently. Code is source of truth; docs are hints.
Phantom Changes in Agent PRs
AI-agent PRs fail at description-vs-diff, not at code. Phantom changes (45%), scope understated (22%), placeholder (19%). Don't trust the agent's own description.
AI in Products
Make Hallucination Visible
Don't promise zero hallucination. Tag every claim as grounded or [UNGROUNDED]. Make failure visible. Match how senior humans work — cite when certain, flag when guessing.
Leadership & Teams
Hiring & People
Portfolio Carry Over Single-Bet Equity
Single-bet equity makes operators fight to stay on winners and abandon strugglers — the opposite of what the studio needs. Portfolio carry aligns them with the studio's return profile.
Communication
The Founder Bottleneck Audit
Two questions at every cycle's end: where did you wait on me, and where was my spec unclear? The first reveals decision-rights gaps, the second briefing gaps.
People Decisions
Stakeholder Bloat
Max three decision-makers per venture. Anyone else is an advisor. Coordination cost grows exponentially with voices, and decisions slow to the most cautious person in the room.
Investing, Markets & Capital
Venture Strategy
Acquirable Beats Fundable
Studios should optimize for acquirable, not fundable. Multiple 5x exits on a predictable cadence beats one 50x lottery ticket while operating burn continues
The LP Education Gap
LPs trained on 2/20 reject studio economics. Bake education into the fundraise, or prove the model first with service revenue, then raise on a track record.
Capital Allocation
The Dual-Entity Structure
Separate the fund from studio ops. The fund hits conventional management-fee math; the studio gets a real P&L; the founder equity story stays clean.
The Management Fee Mismatch
Studios need 10-15% of capital for ops, but standard funds charge 2-2.5%. The cash-flow crisis hits in year two. Fix: dual-entity structures or portfolio service revenue.