At the time, marketplaces were changing the basic economics of a young brand. A small company no longer needed a national distributor or years of retail relationships to reach a large market. It needed a catalog, inventory, reviews, competitive pricing and the discipline to work with a marketplace’s search and fulfilment systems.

That was a large change. It also looks simple compared with what is happening now.

Commerce has moved through three distinct interfaces. E-commerce turned the shelf into a search result. Quick commerce turned it into a local availability problem. AI-driven commerce is turning it into an answer.

Each interface asks something different of a brand.

• A marketplace asks: should this product rank?

• A quick-commerce app asks: should this SKU occupy scarce space in this dark store, in this pin code, today?

• A shopping agent asks: is this the right product for this person, under these constraints, and can I explain why?

The third question is the one founders should sit with. A customer may soon see fewer product pages, fewer grids and fewer brand stories. They may ask for “a mineral sunscreen for humid weather that will not leave a cast on brown skin, under $25” and receive three answers. The brands that make that answer set will have clear product data, credible proof, reliable availability and language that a machine can parse without flattening what makes the product distinct.

That is a different way to build a brand.

What changed after the marketplace era

Marketplaces rewarded operational competence. They made distribution easier and comparison brutal. A founder could reach a country, but every competing price, rating and delivery date sat one tap away.

Quick commerce tightened the loop. In India especially, the digital shelf became physical again. A product can be nationally listed and locally absent. It can have excellent creative and still lose because the right pack size is not stocked within a few kilometres of the buyer.

CareEdge estimated India’s quick-commerce gross order value at ₹64,000 crore in FY25, up from ₹30,000 crore in FY24. It also counted 3,072 dark stores across the three largest operators, 71% more than a year earlier. Those figures are estimates, but the operating consequence is plain: assortment, replenishment and neighbourhood-level demand now shape brand discovery. A 200 ml pack and a 500 ml pack are not merely packaging choices. They are different bets on basket size, price perception, storage space and replenishment speed. CareEdge’s July 2025 report

AI-driven commerce changes the interface once more. OpenAI’s Agentic Commerce Protocol lets a shopping agent pass an order to a merchant while the merchant remains responsible for payment, fulfilment, returns and support. Shopify has introduced agentic storefronts for selling through AI conversations. Google’s shopping work now includes agentic checkout. Amazon says its shopping assistant uses product catalog data, reviews, customer questions, web sources and account memory; it can compare products, add items to a cart and act on price thresholds. OpenAI, Shopify, Google, Amazon

This does not remove marketplaces or stores. It adds another gatekeeper. The gatekeeper reads attributes, evidence, policies, availability and customer history before it recommends a product.

Yesterday, a brand had to be shelf-ready and search-ready.
Now it also has to be answer-ready.

Start the same company in America and India

Imagine two 24-year-old founders launching the same electrolyte drink. One lives in Austin. The other lives in Bengaluru. Both have $50,000, access to the same models and no full-time team.

The product may be identical. The work is not.

The American founder

The American founder enters a mature direct-to-consumer system. Payments, fulfilment, subscriptions and storefront software are easy to assemble. The market is large: the U.S. Census Bureau estimated $1.234 trillion in retail e-commerce sales in 2025, or 16.4% of retail sales. U.S. Census Bureau

Ease of entry creates crowding. The founder has to earn attention across retail media, marketplaces, search, creators, social feeds and AI assistants. Every channel produces its own version of the product truth. The Amazon title, Google feed, product page, creator brief and support answer drift unless someone keeps them aligned.

Returns also belong inside the brand model. The National Retail Federation estimated that 19.3% of U.S. online sales would be returned in 2025. That is a retail-wide figure, not a forecast for electrolyte drinks, but it shows why a founder cannot judge creative on conversion alone. An ad that attracts the wrong buyer can report a strong return on ad spend before refunds, support cost and low repeat purchase expose the mistake. National Retail Federation

For this founder, the important loop joins message, cohort and contribution margin:

claim -> audience -> order -> product experience -> repeat or return -> revised claim

The Indian founder

The Bengaluru founder faces a more uneven market. A premium consumer in Mumbai may discover the drink through an Instagram creator and expect it in 10 minutes. A buyer in a smaller city may find it on a marketplace, compare the price per serving, read reviews in another language and prefer a different payment or pack size.

India’s rails are also more open in places. ONDC describes a network in which buyer applications, seller applications and gateways exchange commerce requests instead of putting every seller inside one closed marketplace. Its public site reported 7.64 lakh sellers and service providers and more than 16 million orders in May 2025. Those are network-reported figures, but they point to a future in which distribution may be spread across many buyer interfaces. ONDC

The Indian founder therefore needs a loop at the level of the city, pin code and SKU:

local demand -> stocked pack -> promised delivery -> fulfilled order -> reorder interval -> local replenishment

The creative system must handle language and cultural context without creating five inconsistent brands. The operating system must know that a Hindi headline, a Kannada support reply and an English product page refer to the same approved claims. It must also keep inventory truth close to the message. There is little value in creating demand for a flavour that is out of stock in the customer’s delivery radius.

America makes media efficiency and post-purchase economics unusually important. India makes local availability, pack architecture and channel-specific economics unusually important. Both founders need the same foundation: clean product data, a durable point of view and a fast learning loop.

So, can a GPU build a better brand than an expert?

Sometimes. Only if we are precise about the job.

A GPU can explore more possibilities than a creative team. It can produce 200 arrangements of a product shot, analyse 50,000 support conversations, translate approved copy into 10 languages, compare hundreds of competitor listings and rerun the work when a price or policy changes.

That makes it good at search: finding useful options inside a defined space.

An expert remains better at defining the space. Which customer deserves the brand’s attention? Which promise can the product keep for years? Which sales lift would damage trust? Which visual code is memorable enough to own and strange enough to defend? A model can imitate the evidence it receives. It cannot be held accountable for choosing the wrong ambition.

The practical answer is:

A GPU can beat an expert at the number and speed of iterations. It cannot decide what the brand should mean unless a person turns that meaning into constraints, examples and tests.

The winning unit is not a model or a creative director. It is an expert who can design a system in which models do the repetitive search, software records the evidence and people retain the consequential decisions.

RLMs matter because a brand has more memory than a prompt can hold

Most brand systems today use a small folder of prompts: tone of voice, audience, prohibited words, visual references. That works for a demonstration. It breaks in a business.

A working brand accumulates too much context: customer interviews, reviews, returns, tickets, ad comments, product specifications, lab results, retail sell-through, creator content, marketplace rules, packaging files and decisions whose reasons matter months later.

Recursive Language Models, or RLMs, offer one way to work across that volume. In the original research, the model treats a long prompt as an external environment. It writes code to inspect parts of the material, breaks the task into smaller calls and combines the results. The authors report strong results on inputs far beyond a base model’s context window. It is a recent research method, not a guarantee that every commercial system will behave the same way. Zhang, Kraska and Khattab, “Recursive Language Models”

The useful idea for a brand operator is recursion, not the acronym.

Suppose 8,000 reviews say a face cream feels “heavy.” A normal summarizer reports that customers dislike the texture. A recursive workflow can do more:

1. Separate verified buyers from other comments.

2. Split feedback by climate, skin type, variant and acquisition channel.

3. Compare the complaint with formulation notes and support tickets.

4. Check whether “heavy” predicts returns, low repeat purchase or only first-week discomfort.

5. Draft different explanations for the product page and support team.

6. test whether the new explanation improves qualified conversion without increasing refunds.

7. Write the result back to the brand’s memory.

Now the model is participating in a learning system. It is not “writing more content.”

The six primitives of an AI-native brand

The tools will change. These six building blocks are more durable.

1. A machine-readable product truth

Create one canonical record for every SKU: ingredients or materials, dimensions, compatibility, certifications, price, inventory, shipping limits, claims, claim evidence, care, returns and approved comparisons.

Store fields as data, not paragraphs. “Works with iPhone 15” should be a compatibility field with a source and last-verified date. A paragraph can be generated from that field; the reverse is unreliable.

This matters because search engines and shopping agents consume structured product information. Google’s merchant listing documentation, for example, specifies fields for price, availability, shipping and returns. Google Search Central

2. A brand constitution

Most brand guides describe fonts and adjectives. An operating constitution records decisions.

It should contain:

• the customer and the problem you have chosen;

• the promise the product can prove;

• claims the brand will never make;

• visual patterns to own and clichés to avoid;

• the acceptable trade-offs between growth, margin and trust;

• examples of good and bad work, with reasons;

• who may approve a price, claim, refund or product change.

Models need boundaries more than adjectives. “Warm, bold and authentic” produces mush. “Never imply a medical outcome; show the product in used, ordinary homes; never hide the cable; quote battery life only from the current test sheet” produces usable work.

3. Generators with constraints

Text, image, video and code models can create product pages, ad variants, retail sell sheets, packaging studies, support replies and local-language versions.

Connect every generator to the product truth and brand constitution. Record the input data, model version, prompt, output and approval. Otherwise, when performance changes, nobody can tell whether the cause was the idea, the audience, the inventory or the model.

4. Tools that can act

An agent becomes useful when it can check inventory, update a feed, create a draft campaign, issue a permitted refund or open a production task. Give it the smallest set of permissions needed. Set spend limits. Require human approval for public claims, pricing changes, regulated advice and large payments. Keep a log.

The goal is not maximum autonomy. It is a short, inspectable path from evidence to action.

5. Evals tied to the business

Before deployment, test outputs against a fixed set of cases. Does the support agent cite the current return window? Does the product comparator distinguish measured facts from opinion? Does a translation preserve the approved claim? Does a generated image alter the product, shade or included accessories?

Then connect quality to business outcomes. Measure contribution margin after refunds, repeat purchase, complaint rate, delivery success and new-customer incrementality. Click-through rate is useful. It is not a brand objective.

6. Memory that updates after a test

Every campaign should leave the system smarter. Record the hypothesis, audience, variants, exposure, result, confidence and decision. Also record what you chose not to learn because the sample was too small or the test was contaminated.

Without that memory, a company repeats old debates with new people. With it, a two-person brand can compound judgment.

Build the loop before you automate it

A sensible weekly loop looks like this:

StageSystem workHuman decisionOutputObserveJoin sales, inventory, support, review and media dataDecide which signal is worth attentionA ranked problem listDiagnoseSegment the signal and retrieve relevant historyChoose a causal hypothesisA testable explanationMakeGenerate copy, images, offers or operating changes within constraintsReject work that damages the brandA small set of variantsTestRandomise exposure, enforce budgets and monitor safety checksStop invalid or harmful testsClean experimental dataJudgeCalculate margin, repeat, returns and uncertaintyDecide ship, revise or discardA recorded decisionRememberUpdate examples, rules and model contextApprove changes to the constitutionBetter inputs for the next loop

The order matters. If a brand automates production before measurement, it creates more untraceable work. If it automates measurement before defining the promise, it becomes very efficient at chasing whatever moved last week.

What this looks like by category

The loop should change with the product.

Fashion: connect creative to size-level sell-through and returns. Generate styling and model variations from approved product geometry, then reject any image that changes fit, colour or construction. Judge the work on kept-order margin, not orders placed. Use return reasons to improve sizing copy and the next buy.

Beauty: connect education to substantiated claims. Retrieve the ingredient record, concentration where publishable, test method, contraindications and customer routine before drafting an answer. Segment results by concern, climate and routine. Escalate medical questions. Judge the system on repeat use and complaint rate as well as conversion.

Food and household goods: model the reorder interval by household or cohort, with inventory and consent rules. Trigger reminders when the estimated need is high, not on a fixed blast schedule. Test recipes and use cases against incremental orders, unsubscribes and waste. In India, make pack size and local availability part of the message-selection logic.

Home: preserve product dimensions, material and colour while generating scenes. Let a customer place a verified 3D or segmented product representation in a real room. Record which room types and dimensions precede purchase. Judge the system on conversion and return reasons such as “too large,” “wrong colour” or “did not fit.”

Electronics: turn specifications, firmware versions, compatibility tables and support history into a cited answer layer. Update comparisons when the source specification changes. Show uncertainty instead of inventing an answer. Always leave a route to a person when compatibility, safety or account access is unclear.

These are narrow systems with feedback. That is why they become part of the company. A one-off image generator does not.

What a young founder should build in the first 30 days

Do not begin with a six-month “AI transformation.” Begin with one decision that repeats often and has a measurable cost.

write the truth. Create the SKU schema, claim register and brand constitution. Import the last six months of sales, support and inventory if the company has them. If the company is new, add 20 customer interviews and the evidence behind every product claim.

choose one loop. In the U.S., that may be ad concept to kept-order margin. In India, it may be pin-code demand to stocked pack and reorder. Pick a baseline and one primary outcome. Add guardrails such as refund rate, complaint rate and gross margin.

run it by hand. Use models for analysis and generation, but make a person approve every action. Keep the batch small. Ten well-labelled variants teach more than 500 assets with no hypothesis.

automate the stable steps. Schedule data pulls. Add evaluation tests. Let an agent create drafts or low-risk actions. Preserve approval for claims, price, spend and customer remedies. Review the learning log at the same time every week.

After 30 days, ask a hard question: if this loop disappeared tomorrow, would the team lose a decision it now makes better? If the answer is no, remove it. Software that produces activity without improving a decision is overhead.

The failure modes are predictable

The first is proxy addiction. A system learns to increase clicks by making the product sound cheaper, easier or more dramatic than it is. Sales rise. Returns and mistrust arrive later. Use delayed outcomes and guardrails.

The second is synthetic sameness. If every brand draws from the same models and the same “best practices,” competence becomes common and distinction disappears. Keep proprietary examples, customer language and deliberate exclusions in the constitution.

The third is false precision. A dashboard gives weak data three decimal places. Require minimum sample sizes, holdouts and confidence intervals. Mark observational findings as observational.

The fourth is broken provenance. A claim enters a feed, spreads to ads and support, then survives after the source changes. Attach a source, owner and expiry date to every consequential field.

The fifth is platform dependence. A marketplace, quick-commerce operator or AI assistant can change ranking, fees or access. Keep the canonical customer and product record in systems the brand controls. Export test history. Build direct permission to contact customers.

The sixth is removing judgment at the moment it matters. Agents are good at frequent, reversible decisions. People should own rare, expensive and reputational ones.

Is anything still the same then?

A brand remains a promise kept often enough that a customer stops reconsidering the choice.

The promise may now travel through a product feed, a dark-store assortment model and an AI-generated answer before it reaches a person. That makes operational truth more important. The package has to match the image. The item has to arrive when promised. The support policy has to survive contact with an unhappy customer.

AI lowers the cost of producing an acceptable version of almost anything. It also raises the standard for what a small team can ship. The durable advantage is the rate at which the company turns real customer behaviour into better product, clearer communication and more reliable delivery without losing its point of view.

Can a GPU build a better brand than an expert?

Give the GPU no constraints and it will produce a large pile of plausible work. Give the expert no system and the company will learn too slowly. Put them inside a measured loop, and a young founder can build with the memory and operating cadence of a much larger company.

That is the new baseline.

They wound down Sora, their photorealistic video AI, widely considered one of the most technically impressive things they’d built.

And they acquired TBPN, a daily tech and business show with 58,000 subscribers and $5M in revenue.

The media covered it like a content strategy.

It’s not.

The real scarcity flipped.

For three years, the AI race was about generation. Better text. Better images. Better video. Sora was the logical peak of that bet.

The problem is that generation scales supply, not value.

The marginal value of the 10,000th AI-generated video approaches zero.

You can flood every channel with content. You can’t manufacture what content shapes decisions.

The scarce resource was never content.

It was always context. What gets believed. What gets repeated. What a VC thinks about when they’re deciding whether to back your competitor.

Subscribe now

What TBPN actually is.

On paper: a media company. Daily show, podcasts, clips, newsletter.

But strip away the media framing, and you have something different entirely.

TBPN is a relationship graph with a broadcast layer.

Direct access to top founders and investors, wrapped in a daily habit loop, delivered through hosts with real credibility built over years.

The data from their show is public. The relationships and the trust those relationships carry are not.

That’s the asset.

And those numbers don’t justify a “low hundreds of millions” price tag unless you’re buying something other than a content business.

The closed loop nobody mapped.

Here’s the architecture I think OpenAI is building toward: Models generate research summaries, question angles, and narrative frames upstream of production.

Human hosts introduce social risk, real-time judgment, and credibility into the distribution layer.

Audience engagement (what gets clipped, shared, discussed, searched) feeds back into model fine-tuning and prompt shaping.

That’s a tighter feedback cycle than anything traditional media has run. And it’s operating on an audience whose downstream decisions carry unusually high-value signals.

This isn’t content production. It’s a closed-loop narrative engine.

Why Sora had to go

Sora produced static artifacts.

A generated video doesn’t adapt after it’s rendered. It doesn’t respond to audience reaction. It doesn’t feed back into the next output in any meaningful loop.

From a systems perspective, Sora was a powerful open-loop generator. High compute cost. No feedback integration. Weak coupling to enterprise workflows, where OpenAI’s revenue actually concentrates.

Cutting it to fund a distribution acquisition with tight feedback properties isn’t a pivot.

It’s a coherent architectural trade-off.

The IPO timing is not a coincidence.

OpenAI is moving toward a public offering.

At that stage, perception compounds faster than any metric. Talent decisions. Capital allocation. Regulatory posture. Enterprise adoption.

All of these move on narrative velocity, setting the first frame of interpretation before analysts, journalists, and short-sellers get to it.

Traditional companies use PR firms and controlled press releases. That model breaks when media organizations are adversarial, news cycles compress to hours, and social distribution fragments across dozens of channels.

Owning TBPN changes the math for starters.

Daily, direct access to the people who move capital. A platform to frame strategy in real time. A feedback channel that detects narrative drift before it becomes a problem.

Not controlling “the media.”

Controlling the first draft.

One thing the automation narrative gets wrong.

There’s a temptation to see this as the beginning of fully AI-generated media.

It won’t be, and the TBPN deal is evidence for that, not against it.

AI smooths edges. Humans create them.

TBPN works because of its hosts. They interrupt, react, speculate, and occasionally get things wrong. Those moments of genuine friction create trust. Trust creates shareability. Shareability is what makes the distribution layer function.

The emerging model isn’t AI replacing human media.

It’s human judgment amplified at an infrastructure scale.

What comes next?

AI labs will acquire or build media properties. Distribution channels will become model-aware. Small, high-signal audiences will command premiums that their ad revenues dramatically understate.

Daily loops will outperform static content libraries.

Narrative speed will become a measurable competitive advantage, not a PR metric, but a structural one.

The long-run implication is worth sitting with.

If intelligence becomes the primary interface people use to make sense of the world, and distribution is embedded within that interface, the entities controlling both layers don’t just shape what people see.

They shape what people consider worth seeing.

Subscribe now

Phew! It’s been a while since I wrote one of these. But now I’m back! So, heads up, world!

What am I doing now?

I’m building ShopOS, that offers an army of agents to create, market, and manage your ecommerce brands.

Will share more about what we do in the following emails.

It has been a minute, a pandemic and then some. To dial your memory, my team and I were building the worlds first No-Code platform for making Virtual & Augmented Reality experiences. We called it Scapic, mostly because the domain was available and partly ended up being a word that everyone made a mistake on.

We started Scapic with a goal to be the simplest way you create, share and explore AR, VR & 3D experiences. Since then, we've seen hundreds of thousands of experiences generated on Scapic's No-Code platform. Towards the end of last year, Scapic was acquired by Flipkart.

A few things have changed since then, but the mission to bring Augmented Reality to a billion Indians, is now stronger than ever.

The journey of building Scapic, or any deep-tech startup is not for the faint hearted. If any of your friends tell you that they are going to build a deep-tech startup, the biggest favour you can do is to shoot them, while they are still happy.

If they still persist, send them this newsletter because it’s meant to help with tactical insights on what it takes to build right on the frontiers of tech. I’m writing Battle Cards to be the stash of information that I wish I had, when starting up.

Share Battle Cards

The journey so far

Ever since Engineering undergrad, I’ve been trying to make things.
Most don’t work out, but some do. Looking back over 8 years of trying to launch products, 3 unique phases stand out, and I documented each of them:

📗 Startup 1: Aeriem:
Adaptive learning LMS for K-12 education

🛵 Startup 2: Supertext:
WhatsApp based hyperlocal delivery

🥽 Startup 3: Scapic:
No-Code platform for Augmented Reality

Every journey taught me something new, something that changed how I see the world, and pave the way further ahead.

A shiny new website

We do not have enough software developers in the world, and in turn, great quality software. The No-Code revolution is something I’m deeply excited about for changing this. So, decided to launch thesaitec.com as a site to capture the things I do, the startups I advise and what I’m passionate about.

Check it out

Launching Battle Cards

Startups = speed of execution
Faster execution = more chances of mistakes
To Avoid mistakes = read and learn

Battle Cards is a newsletter about ideas, principles and actionable insight into what goes behind building deep-tech products. Here is a quick post on what Battle Cards are, and why build it:

Why build Battle Cards?

Initially, I experimented with screenshot sized essays that are optimised for mobile reading. Whilst the content resonated, the format, seemed better suited to be a newsletter format

One of the first Battle Cards I wanted to write out was a map of reflection that I found useful as a founder, after Scapic’s acquisition. I think a good map of reflection is incredibly useful, and wrote more about it here:

Map of Reflection

Launching things in 2021

This year, my goal is to better document my learnings as an entrepreneur, builder and product geek. There are several launches that I’m deeply excited for. Through it all, I’d look to document my learnings often using Battle Cards. Adjust your veshtis as we’re just getting started.

Subscribe to Battle Cards

Lot’s done.
Lot’s more to do.

☕️ & 🥞,
Saitec

Inertia

A lot of why we beat ourselves up is due to the lack of forward movement or inertia. It’s dangerously easy to fall into this state, and not as easy to shake it off. A few pointers helps:

• Get out of the building: Whatever it is that you are upto, stop confining it to just the walls of your cabin or your living room. Step out, meet a wide range of peers or better, leaders. Understand what they’re about and how they go about their days, habits and lead their teams.

• Stop looking for role models: Instead, keep an eye out for individual traits. Things that you want to develop, in your way, voice and style. 

• Appoint a board of directors: You need a personal board of directors, or people who can root for you in specific walks of your life! These would be the people you can rely on for advice and feedback. Don’t look for answers from them, but better ways for you to think to come up with the answers. 

• Your support crew at work: Your manager can appear as a hostile superior, but flip the power dynamic. Think of your manager as a partner in your journey who can help you better if you articulate clearly where you want to go, and why that’s the right thing. 

• Set goals: Self advocacy is incredibly useful to help identify growth areas and help articulate it to the people who can help you in the journey

Draw out a map!

Self reflection is often a scary and ambiguous exercise for a lot of us. With a map, maybe it doesn’t have to be that way. The map is built upon 3 dimensions to help pull apart different areas to explore:

1. The Scope: Concentric circles from the center. For example, if you’re a founder, your role has 3 scopes of impact: 
   The Project
   The Team
   The Company
   Items can be listed across these scopes, and live on the boundaries too. 

2. People vs Products: This makes your x-axis. 
   We tend to focus on the products we build. 
   However, being an effective teammate is much more than that. 
   The impact on people is as important as the launch.

3. Past vs Future: This makes your y-axis. 
   The top half would be the last 6 months
   The bottom half would be the goals for the next 6 months. 

Let’s begin with a rewind

Pick up the breadcrumbs. Review every surface of productivity you have. The calendar, slack chats, todo lists, project Management tools, coffee meetings, 1:1s, e-mails (every one of them)

Now, ask yourself:

• What did you do?

• Why was it hard? 

• What did you learn?

💖 ⭐ 🚫 Map

Plot the top half of your map now, and ask yourself: 

• What was your role for each item?
  Label as: learning, doing, leading, or teaching

• What did you love doing?
  💖 those items 

• What are you uniquely good at (as told to your teammates)?
  ⭐ those items

• What did you hate doing?
  🚫 those items

💎 Identify Themes

Now, you have the basics to identify themes. You may see the things you are most comfortable in, or take the most amount of joy in doing. Take this to your personal board of directors and see if they agree with what you’ve discovered about yourself. 

With this information, craft out your narrative. This is almost an elevator pitch for just you. 

• I am strong at …

• I am the go-to person for …

• I am at my best when partnered with …

What’s ahead

Predicting the future is hard, but improving yourself to take it on, need not be. 

Set a foundation: The bottom half of your map is now for the next half. Transfer the things you plan to continue

Identify opportunities: Growth lies in doing new things. Add any new projects, interests that align with your narrative. f you have an empty quadrant, try something new! What is something you’d like to learn or someone you’d like to learn from? Ask the people around you for ideas or collaborations.

Double down: Your narrative holds what you are strong at. How can you expand on it? Double down on your strengths. 

Rebalance: It’s easy to add dozens of new oppourtunities, but only to feel overwhelmed about it. Now, go back to your personal board of directors on what’s to prioritize and rebalance out. 

Once done, filter your near final 6 month map with a few questions:

• What fuels you?
  Star the things that you will continue to do because they give you energy and sustain you.

• Where are the gaps that your company/team needs filled?
  Talk with your manager about your unique role. Star the items that no one else can do.

• What are the things you can deprioritize for now?
  Cross out items you enjoy, but need to scale back or postpone, to avoid getting stretched too thin.

Gaining Support

Once you have a clear vision of where you want to focus next, then it’s all about followthrough. Co-conspirators and accountability buddies are key. Share your goals and intentions with your friends. Find people who have similar goals and go after them together. Draft it into a goals doc and share it with your manager, your team, and your peers. Who knows, you might inspire someone else to think about their own goals a little differently.

Reflection has helped me immensely to better plan, understand and be deliberate about my time, priorities and confidence in building things.

Startups = speed of execution
Faster execution = more chances of mistakes
To Avoid mistakes = read and learn

Reading more is not the same as being better. Applying your learnings is key. If you do not know how to join the dots, you're just left with the canvas, no matter what.

I began to note down the key learnings from everything I read that I found useful. Through the coming decade, I seek answers to one profound question:

How computing, content & commerce might evolve in Virtual Reality & Augmented Reality

Ever since the acquisition of my startup, Scapic, I’ve been looking at how immersive technologies will shape the future. There is a lack of pointed, tactical advise for founders, entrepreneurs and product managers operating in the Deep-Tech space.

Battle Cards is a newsletter about ideas, principles and actionable insight into what goes behind building products on the bleeding edge. The goal is to highlight the non-obvious things that would help you build better.

Share Battle Cards

Specifically, this newsletter will include…

• Science Fiction: Explainers on how Extended Reality will change our world

• Productivity: Screenshot sized essays on how do you build the future as a high agency team

• Ideas: The next big thing will begin out as a toy, before it disrupts an industry

  

⚔️ Join the horde!

Let’s fall down the rabbithole, together.
Sign up now so that you don’t miss a single battle card.
They’re built to be concise, timeless and hopefully, interesting too.
The goal is that somewhere in here is an idea that could change what you’re building.

Subscribe now