Silicon Valley GirlSuperhuman CEO: How to Position Yourself Now Before the Next AI Phase (2026–2027)
CHAPTERS
- 0:00 – 2:56
Stand out by avoiding the recruiting funnel: create visible work
Shishir explains that the best opportunities often come from non-recruiting interactions, not from applying through formal pipelines. To get into the “interesting people” pile, he recommends doing public, high-signal work that naturally attracts conversations.
- •Most career transitions happen through organic interactions, not job applications
- •Recruiters’ inboxes become crowded folders; you want to be in the other pile
- •Be interesting in the world: start projects, write, publish, ship
- •Distribution is easier than ever; work can spread via networks, not just social media
- •Example: a paper (“Four Myths of Bundling”) led to Spotify board-level conversations
- 2:56 – 4:16
Why “learn every AI tool” misses the point: learn to manage
Instead of chasing every new AI product, Shishir argues the differentiator is learning management-like skills. AI tools put more people in a “manager of execution” role earlier, shifting what strong performance looks like.
- •Generic advice to learn every tool is less valuable than building durable skills
- •AI turns individual contributors into workflow/tool/agent managers
- •Execution skills still matter, but orchestration and decision-making matter more
- •The world is reversing: managing is becoming a more universal baseline skill
- 4:16 – 4:26
What to do when manual work is automated: companies hire judgment
Marina challenges how people can develop good instincts if the entry-level “manual” phase disappears. Shishir responds that judgment is exactly what organizations are hiring for, and it must be built through deliberate practice.
- •Automation removes some repetition, but raises the premium on judgment
- •Hiring focus shifts from “can you do the task” to “can you decide what’s right”
- •Judgment is learnable, but it requires repeated reps and feedback loops
- •The key is designing practice opportunities even when work is high-stakes
- 4:26 – 6:46
Practice in low-stakes environments to build real skill
Shishir introduces the idea that skills develop best in low-pressure settings, like practicing sports or music. He recommends side projects and rapid feedback loops to learn AI-centered design and judgment before high-stakes workplace use.
- •People mistakenly try to learn in the highest-stakes situations
- •Analogy: you don’t learn guitar only through recitals—practice privately first
- •Use side projects/friends to create fast repetition and feedback loops
- •Find the “driveway hoops” equivalent for your job to practice safely
- 6:46 – 7:55
Career ladders aren’t erased: PSHE and how promotions actually work
Shishir outlines his PSHE framework—Problem, Solution, How, Execution—to explain seniority. Promotions shift from executing assigned tasks to defining problems and spaces, and success is not just about scope.
- •PSHE ladder: from executing instructions to defining problems and spaces
- •Early: given problem/solution/how; you execute
- •Mid: you figure out the “how,” then create solutions
- •Top: you identify the right problems in a broader space
- •This ladder is an axis separate from “scope” and project size
- 7:55 – 9:02
The “trough of disillusionment”: why mid-career promotions feel confusing
Using an S-curve model, Shishir explains why people hit a mid-career moment where bigger scope stops being enough. At that stage, evaluation shifts toward how you operate—problem framing, solution quality, and leadership mechanics.
- •Teams can be mapped across two axes: scope and PSHE maturity
- •Early careers: grow mainly by scope while staying execution-focused
- •Mid-career: evaluation flips toward how you do the work (the trough)
- •Promotion committees compare same-scope people by judgment and approach
- •Applies across functions: engineering, design, sales, marketing
- 9:02 – 13:13
AI shifts the ladder upward—and why he avoids the word “replaced”
Shishir argues AI doesn’t eliminate the ladder; it supplies “great executors” that raise what humans can build. He reframes the fear of replacement as a zero-sum narrative and uses historical tech analogies to show work expands into bigger ambitions.
- •AI can assist at every PSHE layer, but humans must judge and choose
- •Problem selection and framing remain predominantly human responsibilities
- •“Replaced” implies zero-sum; Shishir prefers a growth/expansion frame
- •Analogy: power tools didn’t remove construction—enabled skyscrapers
- •Result: broader capability means building bigger things with the workforce
- 13:13 – 15:42
What happens to entry-level roles: taste, strategy, and scalable creativity
Marina describes hiring juniors who can’t just draft content because AI can already do that. Shishir argues the underlying creative skill still matters—great ideas and taste become more valuable because AI scales them and multiplies variants.
- •Entry-level execution shrinks; expectations move toward strategy and taste
- •Marketing analogy: shifted from one perfect ad to millions of variants
- •Original creative skill doesn’t disappear; it’s elevated and scaled
- •AI amplifies great ideas; humans supply the spark and the judgment
- •New bottleneck: reviewing, selecting, and steering outputs at volume
- 15:42 – 17:00
Eigenquestions: the most discriminating skill he hires for
Shishir introduces “Eigenquestions,” a framework for identifying the question that unlocks the most other answers. He connects it to PSHE and calls it his top hiring signal, emphasizing it’s learnable through practice in safe settings.
- •Eigenquestion = the question that, when answered, resolves many others
- •Hard part isn’t the answer—it’s asking the right question
- •Maps to PSHE: Problem (question) precedes Solution/How/Execution
- •He trains and recruits for this; it’s his #1 indicator for hires
- •Best learned via repeated practice in low-stakes environments
- 17:00 – 19:39
The teleportation interview: how great question-framing works
Shishir shares a favorite interview exercise: bringing a teleportation device to market. Candidates must narrow many questions down to two, revealing their ability to choose high-leverage uncertainties and build a go-to-market approach.
- •Interview prompt: “Teleportation exists—how do you commercialize it?”
- •Candidates list many questions, then must choose only two
- •Example Eigenquestion axis: safety for humans vs not safe
- •Second axis: CapEx vs OpEx shapes deployment model (airports vs everywhere)
- •Shows how framing creates actionable strategy without needing perfect info
- 19:39 – 22:09
How to practice judgment without risking your job
Shishir recommends turning Eigenquestion practice into games with friends and exploring many hypothetical domains. The goal is to build the muscle of identifying decisive questions without workplace stakes or reputational cost.
- •Use playful prompts; even kids can surface the core question quickly
- •Create a personal list of scenarios to practice repeatedly
- •Do it socially for fast feedback and diverse perspectives
- •Practice builds big-picture thinking rather than reactive problem-chasing
- 22:09 – 24:03
Inside Grammarly’s scale: the ‘AI superhighway’ and 100B queries/week
Shishir explains Grammarly’s core advantage: bringing AI into the exact surface where people work, at the speed of typing. He shares surprising scale stats and reframes Grammarly as infrastructure for many agents, not just grammar correction.
- •Grammarly isn’t just grammar—it’s AI embedded where you work
- •Scale: 100B LLM queries per week; ~3,000 per user per day
- •Runs across ‘a million unique surfaces’ (apps, web, desktop, mobile)
- •The infrastructure is an ‘AI superhighway’ that can carry many agents
- •Grammarly was one ‘car’ on the highway; now it can be generalized
- 24:03 – 28:08
Demo: building ‘Assist’ agents that work while you work (fact-check, legal, PM)
Shishir demos Superhuman Go, an agent builder that triggers while writing and surfaces proactive underlines inside your tools. Examples include fact-checking, launch project status, and legal guardrails—without switching windows or remembering to ask.
- •Agents have tools/connectors (mail, calendar, docs, MCPs) and instructions
- •Key innovation: ‘while writing’ trigger for proactive suggestions
- •Examples: launch PM agent (status), legal guardrail agent (compliance language)
- •Fact-checking can pull context from your actual data sources automatically
- •Vision: choose who sits on your shoulder, beyond a grammar teacher
- 28:08 – 33:59
Chat vs Do vs Assist: the AI interaction model shift + top daily agents
Shishir contrasts three AI metaphors: Chat (conversation), Do (task completion), and Assist (ambient proactive help). He shares his most valuable agents—fact checker, placeholder filler, and calendar-aware scheduling—and why proactive AI avoids becoming a chore.
- •Chat: you ask; Do: it executes tasks; Assist: it helps before you ask
- •Assist drives far more interactions because it’s embedded in the workflow
- •Top agents: fact checker; placeholder filler (fills bracketed gaps); calendar helper
- •Assist reduces context switching and prevents errors early in the process
- •Customization is prompt-driven: you can define behavior like ‘detect double-booking’
- 33:59 – 38:48
Bill Campbell’s lesson: measure success by other people’s success
To close, Shishir recounts advice from legendary coach Bill Campbell: optimize for helping others become successful. Campbell’s personal metric—how many mentees became Fortune 500 CEOs—reframed leadership as non-zero-sum and deeply people-centered.
- •Campbell coached Shishir on core CEO tasks: meetings, firing, hiring, investors
- •He declined compensation, saying his gains went to charity and company growth mattered more
- •His motivation: a numeric metric of mentees who became Fortune 500 CEOs
- •Leadership means rooting for your people, even when they move on
- •Takeaway: define a metric for the success you create in others
