ThredUp (an online consignment & thrift store) carries 35,000 brands across 100 categories, so inventory is constantly flowing in from individual sellers who want to get rid of clothing. But the issue users that shop on ThredUp are facing is how to sift through that overwhelming inventory. ThreadUp, and most resale platforms, hand you a search bar and wish you luck. The problem at hand has always been discovery. Whether you’re in a thrift store or on a digital second hand marketplace, you’re gonna experience fatigue when it comes to picking out the pieces that are right for you.

ThredUp’s Shop Social feature helps with this. This feature aims to prove that the best search is no search at all, because the clothes should come to you.

What Shop Social Actually Does

It’s pretty straight forward; connect your Pinterest account, and ThredUp crawls your boards in the background every time you open the app. It pulls your saved pins, reads your visual preferences, and maps them against ThredUp’s live inventory in real time. The result is a personalized secondhand feed built entirely from the aesthetic you have already been curating, without a single search term required.

This is a meaningful shift for a platform that has historically been search-heavy. Platforms like Depop have long understood that discovery-based browsing drives engagement because search requires intent (and people are lazy lol) but discovery takes away the hassle. Shop Social moves ThredUp closer to that model, and Instagram and TikTok integrations are already in the pipeline.

The Tech Behind It: Two Layers Working Together

Shop Social works because two separate systems are doing their jobs well and communicating with each other.

The first is Pinterest’s Taste Graph. Every pin you save, every board you build, every visual interaction you have on the platform feeds into a behavioral map of your aesthetic preferences over time. Pinterest’s multimodal AI processes combinations of text, images, and voice to understand style intent, not just surface-level keywords. The Taste Graph produces recommendations with relevance scores more than 30% higher than traditional models, because it’s working from a deeply personalized signal rather than a generic query.

The second layer is ThredUp’s own inventory infrastructure. ThredUp has spent years building AI-powered automated digital measurements and item tagging at scale, which is genuinely difficult in resale where every item is essentially one of a kind. CEO James Reinhart has described each item ThredUp sells as a snowflake: the platform might have a handful of black Lululemon leggings at any given time, all in different conditions, sizes, and price points. Getting the tagging right is what makes matching possible. Image search sessions on ThredUp already result in 85% higher conversion rates than standard sessions, which tells you the infrastructure is working.

Shop Social takes Pinterest’s style signal and runs it through that tagging layer to surface inventory matches in real time. It is not a recommendation engine guessing what you might like. It is a matching engine working from data you already curated yourself.

Subscribe now

Why This Is a Bigger Deal Than It Looks

Reinhart has said ThredUp’s AI investments have been “underhyped”, and the numbers back that up. AI-driven improvements to search have produced an 18% lift in new customer conversion and 32% year-over-year growth in new customer volume. Those are significant numbers for a platform operating in a crowded market.

The deeper point is behavioral. When discovery stops feeling like a chore, resale stops being something you do occasionally and becomes something you check habitually (like Pinterest!). That habit formation is what every resale platform is chasing, and it is very hard to manufacture through search alone.

The Verdict

Shop Social is the most technically interesting thing happening in resale discovery right now. The secondhand apparel market is projected to reach $367 billion by 2029, and the platforms that figure out discovery first will have a structural advantage. Pinterest users are already doing the aesthetic curation work. ThredUp is just making it shoppable. The real unlock comes when the Instagram and TikTok integrations arrive and trend-driven signals get layered on top of Pinterest’s mood boarding. When that happens, the gap between seeing something you want and finding it secondhand gets very small and ultimately pushes us all towards sustainable consumption.

What Croissant Actually Does

The product is a Chrome extension and mobile app that shows guaranteed buyback prices on items from 100+ retail partners before you check out. Partners include Nordstrom, Revolve, Mytheresa, Reformation, Bergdorf Goodman, Neiman Marcus, Moda Operandi, and COS, among others. When you’re ready to sell, just click the button and a prepaid shipping label will be ready for you to download. Croissant buys the item back, gives you credits to spend within the network, and resells the goods via its own storefront or platforms like Poshmark, The RealReal, and Depop.

The model is closer to a financial product than a resale marketplace. Croissant takes a 15% cut of each transaction and guarantees buyback at 30 to 40% of MSRP (Manufacturer’s Suggested Retail Price). On a $300 pair of jeans, that’s up to $120 back in spendable credits. CEO John Howard came from private equity at KKR, and it shows in how the product is structured. This isn’t purely a secondhand shopping app, it’s more like a liquidity layer built on top of retail.

Subscribe now

The Numbers

Croissant just closed a $28M raise in February 2026, bringing its total to $52M. The round includes $14M in equity led by existing investors Portage, Third Prime, and George Roberts, alongside $14M in debt capacity. To date the company has driven over $50M in GMV (Gross Merchandise Value) across 100K+ users. The $250M GMV target for 2026 represents a 5x jump from where they sit today. Howard projects profitability by year end.

The new funding is being deployed two ways: acquiring new shoppers on behalf of brand partners, and pre-funding demand by paying brands upfront for purchases Croissant predicts its users will make. In exchange, Croissant takes a bigger cut of each transaction. The company is also now sharing affiliate revenue directly with shoppers, offering 10% back in credits on every purchase made within the network. Average order values in Croissant carts are $700, which means the 10% credit translates to roughly $70 dropped into a spendable balance every time someone shops.

Where The Model Gets Complicated

Their model makes total sense with items on the higher end. Like a Loewe bag, a pair of Agolde jeans, or a Toteme coat; all of which hold resale value in a way that makes the Croissant proposition justifiable.

The harder question is what happens at mid-market. When the 30 to 40% buyback guarantee starts to look less like a financial benefit and more like a small reward for an inconvenience. Whether the new 10% instant credit back is enough to drive habit formation at scale is unclear. Howard has acknowledged the original model did not offer enough instant gratification for the shopping audience, which is partly why the affiliate credit share was added.

The unit economics tension is also very real. Croissant is simultaneously pre-paying brands for future demand, managing resale logistics, guaranteeing buyback prices, and splitting affiliate revenue with users. Each of those is a cost center, the model requires high average order values and high retained resale values to hold together. This means that the higher-income customers are kind of bearing the load for the entire business. That customer definitely exists, but the cohort is not large. What acquisition looks like outside of that group is the central question that the new $28M funding round is trying to answer.

The Verdict

Croissant has built something genuinely useful and even revolutionary to how we shop. The framing of fashion as a liquid asset rather than a sunk cost is smart and the timing is right because the secondhand apparel market continues to grow (projected to reach $350M by 2028). The 5x GMV target and profitability projection by year end are ambitious claims worth watching closely. If the unit economics hold at scale and Croissant can acquire users beyond its natural investment shopper base, this could be the most interesting business model in fashion resale right now.

What’s Different

So, how is Claybird doing this? The answer involves a technically complex pipeline. Most AI ad tools hand a prompt to a video generation model and just hope for the best, but Claybird meticulously runs a full agentic workflow. It starts with a single image NeRF, a neural radiance field that reconstructs 3D geometry from a single product photo, so the model actually understands what the object looks like from angles it was never shown. Fine-tuned image editing models then maintain visual consistency across shots. Video generation runs through top models including Veo and Sora. And lastly, an eval layer checks output quality before anything gets delivered, looping rejected frames back for regeneration rather than shipping them to the client.

YC described it this way: Claybird functions like an expert director, understanding which camera angle, shot type, and lighting each scene requires, and knowing which image and video model to use to get there.

The result is studio quality video creative in under 72 hours, at a fraction of what traditional production costs. For context, a standard agency video shoot involves coordinating actors, crews, revisions, and other various bottlenecks; taking anywhere from weeks to months. Creative gets filtered through layers of people, and by the time the ad ships, the culturally relevant window has often already closed.

Take a look at this Holiday ad Claybird made for Coca-Cola:

Subscribe now

They’re Growing Fast

In their first month, Claybird signed $67K from brands including Coca-Cola, Eight Sleep, Delve, and Mercor. The appeal is not just speed and cost, it’s personalization at scale. Traditional production limits marketers to three to five audience segments. Claybird customers have expanded from two to twelve or more targeted cohorts in their first week. When your cost per creative drops, you can stop running one campaign for everyone and start running many campaigns for the people who actually convert.

However, a concern I have is what gets lost to tech of this caliber and scale. An AI pipeline that can produce twelve variations of a campaign in the time it used to take to produce one is genuinely powerful. It’s also, if deployed carelessly, a machine for volume over voice. The brands that use Claybird well will be the ones that bring the creative direction themselves and use the pipeline to execute it at scale. The ones that outsource the thinking entirely will still end up with slop, not better work.

What This Means for Fashion

A couple weeks ago I wrote about AI moving toward the runway: AI models in campaigns, generative backdrops at fashion week, digital twins replacing human talent in e-commerce shoots. Claybird is a different entry point into the same story. It’s not necessarily coming for the runway, it’s coming for the campaign budget, the performance spend, the TikTok and Reels creative that fashion brands are producing constantly.

Writing about these things is not the same as endorsing the erasure of craftsmanship. It’s accepting that the shift is already happening and that understanding tools is better than being surprised by them. Because, like it or not, change is inevitable.

The Afghan coat is a good place to start. Long before the 1970s hippie trail made it a Western fashion statement, the coat was a staple of Afghan daily life, rooted in the practical needs of a mountainous climate. Shepherds and rural communities wore sheepskin and shearling coats for warmth, and over time the hand embroidery that decorated them became a form of artistic and cultural expression, with patterns and techniques varying by region. When Western travelers moved through Afghanistan along the hippie trail in the 60s and 70s, they brought the coats home, and the garment was adopted by counterculture fashion without any context or credit. What the brands below are doing is reclaiming that context and giving it back.

Aql & Nafs is a London-based label founded by Afghan twin sisters Zohal and Mursal. Their Heritage Collection is handmade entirely by women artisans in Afghanistan using embroidery techniques passed down through generations. Every piece is one of a kind. They’re also explicit about their mission: to bring visibility and economic opportunity back to the original makers, at a time when Afghan craft is increasingly being reproduced in neighboring countries without attribution.

Subscribe now

Zeba Threads is based in Barcelona and specializes in Afghan suzani coats, produced in partnership with artisans in Afghanistan. Suzani is a form of embroidered textile that has roots across Central Asia, and Zeba Threads positions it within a contemporary silhouette without losing the density and detail that makes the craft unique.

Nukra’s pieces are built for longevity and are rooted in practicality while still showcasing the beauty of Afghan patterns to endure both time and trends. Alongside Afghan coats, Nukra produces kurtis and other regional garments as outerwear, keeping the range grounded in everyday Afghan dress rather than treating heritage as purely ornamental.

**A couple of these brands were found using SSQRD, a search platform built specifically for sustainable and ethical brands. If you haven’t used it yet, it’s definitely worth bookmarking.

What Phia Does

The core idea is genuinely useful. Phia is a free iOS app and Chrome/Safari browser extension that functions as a real-time price comparison tool for fashion, much like Google Flights. You browse as you normally would, hit the “Should I Buy This?” button, and Phia instantly scans 40,000+ retail and secondhand resale sites to tell you whether what you’re looking at is overpriced, fair, or a good deal. If the price is bad, it surfaces cheaper exact matches and similar alternatives on the spot. Behind the scenes, it runs on a database of 350 million secondhand items, one of the largest in the US, and uses patented AI to compare price, resale value, and visual similarity simultaneously. It also learns your preferences over time, factoring in your size, brand loyalty, and shopping habits to strengthen its recommendations.

The business model is affiliate-based. When a brand makes a sale through Phia, the app takes a cut, and the company says it delivers a 15% increase in average order value and 30% stronger new customer acquisition for its 6,200 brand partners. The longer-term vision is a full shopping agent; a digital closet that catalogs past purchases, syncs to your calendar to suggest when to buy and when to resell, and handles the entire purchase decision from discovery to checkout.

Subscribe now

The Metrics That Matter (and the Ones That Don’t)

The numbers look strong on the surface, but are worth a closer look. 750K downloads is not the same as 750K active users, and Phia has not disclosed any monthly active user figures. 11x revenue growth since launch is impressive, but the baseline is never mentioned, and 11x of a small number is still a small number. The company is also operating one of the most effective founder-led marketing machines in the fashion tech space right now. Gates and Kianni have a podcast (The Burnouts) on Alex Cooper’s Unwell Network, a combined social following built precisely for their target demographic, and a press cycle that has barely paused since launch. That kind of distribution is genuinely valuable, but it also artificially inflates top-of-funnel numbers in a way most startups can’t replicate and most investors don’t fully account for. The real question is what retention looks like (a metric not disclosed to us) when the novelty wears off and users aren’t being nudged by a podcast episode.

Which brings us to the product’s actual performance. Because the concept and the execution are two different things. Independent testing has flagged lag during searches with high result volumes, which makes sense given the backend is scraping 40,000+ sites in real time, but lag in a price comparison tool is a retention killer. More critically,user reviews flag that the results simply aren’t reliable: one Trustpilot reviewer noted that digging around online independently surfaces better options than Phia does, and that major retailers are missing from results entirely. The app’s affiliate model means Phia has a structural incentive to surface partner results, not necessarily the best results. If a regular Google search consistently returns more accurate price comparisons, users will notice, and they will stop opening the extension. Another thing to note is that the Apple Store reviews for Phia seemed skewed, and most likely fake…

The Privacy Problem

This is the part most coverage has skipped entirely. In November 2025, security researcher Maahir Sharma discovered that the Chrome extension was logging the complete HTML of every page a user visited through a function buried in the code called logCompleteHTMLtoGCS, compressing it, and uploading it to Phia’s servers. Every page. Not just shopping sites: Gmail inboxes, bank statements, Instagram DMs. After Sharma contacted Phia, the company quietly removed the feature without disclosing what had happened to users or confirming what had been done with the data already collected. A separate NowSecure analysis found the iOS Safari extension had nearly identical capabilities. Phia’s official response was that the data was aggregated, anonymous, and never stored. The researchers who reviewed the findings disagreed.

For a shopping agent, this is a structural problem. The entire value proposition of a tool like Phia depends on users trusting it to live inside their browser. Phia had that trust, burned it quietly, and hoped no one would notice.

The Verdict

Phia has real product-market fit, the problem is not the idea, it’s that the hype has outpaced the accountability. A $185 million valuation for a company that is less than a year old, has never disclosed active user figures, quietly buried a data collection scandal, runs a growth strategy built heavily on its founders’ personal platforms, and has real questions around result reliability is a bet on the story as much as the product. The founders who win in the AI shopping space long-term will be the ones who earn trust at scale. Phia still has a real shot at being that company. But it has some trust and consistency to rebuild first.

The reality is that the tech is here, and it’s clearly not going away. But the real question is how far it’ll go, and what it means for the raw, real, and meticulous artistry that fashion week has always represented.

Campaigns are the First Target

AI’s entry point into fashion has been campaigns, not runways (yet), and brands have been moving fast. H&M announced plans to create 30 “digital twins” of real models. These “clones” would possess the same likeness with infinite outfit variations and no reshoots needed. By July 2025, those first images were live: photorealistic digital clones of real models posed against city backdrops, wearing denim, watermarked as AI. H&M framed it as a collaborative effort; models retain rights to their digital counterparts and are compensated according to existing agency rates. Think of it as MAAS: Modeling as a Service. But critics were quick to point out that the photographers, stylists, and makeup artists in the equation have no equivalent protection.

Furthermore, Mango, Jil Sander, MCM, and Burberry all ran AI-assisted campaigns in 2025, ranging from AI generated environments to digitally extended shoots. Hugo Boss went further, launching fully AI-powered product content across its global e-commerce platforms. The brand noted it had been digitizing its product development processes since 2023, with a goal of over 90% digital production by 2025. The economics make it hard to argue against: AI generated e-commerce imagery can reduce production costs by up to 70% compared to traditional studio shoots, with campaign work delivering savings closer to 50%.

For luxury brands, most are still treating AI as an "add-on" by using it to extend existing shoots rather than outright replacing them. The Valentino backlash became a cautionary tale to other luxury houses. Even though their post was clearly labeled as AI, it didn't help. Dr. Rebecca Swift, SVP of creative at Getty Images, said in response to the backlash that consumers hold expensive brands to a higher standard. Being fully transparent about AI usage does not exempt them from public expectations. The main issue isn't the technology, it’s what its replacing.

Subscribe now

What Would Actually Change About Fashion Week?

Campaigns are one thing, but the runway is a different conversation.

The physical fashion week show is expensive, logistically brutal, and produces content that lives online for maybe 48 hours. Generative video tools like Runway and Sora can now produce photorealistic walking figures. The technical gap between “AI generated model” and “convincing enough to publish” is closing faster than most people expected a year ago.

At NYFW 2025, Alexander Wang used AI generated graphics for the backdrops of his Spring 2026 “Matriarch” collection. No models were generated, but AI was woven into the show experience. He also debuted a 3D-printed, zero-waste stiletto created using machine learning. Wang was direct about his stance: “I don’t believe AI is going to replace creativity; creativity is the only disruptor.”

Meanwhile, the INARA fashion show in Kochi, India ran an entire VR runway built with generative AI and motion capture. It consisted of no physical samples, no venue, no shipping. Digital models walked a fully rendered runway in real time while perfectly mirroring human movement. Kind of scary isn’t it?

The argument for going virtual is straightforward; reach millions globally, eliminate venue costs, skip sample production and international logistics, and make fashion week accessible to people who will never receive an invitation.

But fashion week isn’t just content, it’s also buyer previews and press access. It’s relationship building, cultural moment-making, and the kind of intentional, intricate artistry that doesn’t transfer to a render. The show speaks to the fact that a designer cared enough to build something real, something you have to be in the room to fully feel.

The Human Model Problem

This is where it gets genuinely complicated.

The New York State Fashion Workers Act went into effect June 19, 2025, and it’s a meaningful step. Brands and agencies are now legally required to obtain explicit written consent before creating or using a model’s digital replica. The act closes a real loophole, but it also reveals how far behind the regulatory framework still is. Models are classified as independent contractors, which means they can’t unionize and have far fewer protections than Hollywood actors, who negotiated AI rights directly into their SAG-AFTRA contracts after a full industry strike.

Research from Cornell’s Worker Institute and the Model Alliance found that AI has made “Frankensteining” dramatically more common. Aka the practice of stitching together body parts from multiple models without consent or payment. A brand might hire a contractor to take a single approved photo and use AI to generate a dozen outfit variations, basically turning one paid shoot into an entire catalog. The worst part is that the models in those images often have no idea.

The reaction to Gucci’s Milan Fashion Week AI campaign put a finer point on it. One commenter captured the sentiment that surfaced repeatedly: “Why should people pay for a luxury brand that doesn’t even respect itself enough to use real people?” That’s not just a consumer complaint, it’s a cultural signal about what people actually want from fashion at this level.

And designers are responding. On the FW 2026 runways, an emerging counter-trend has taken shape: raw edges, hand-drawn prints, visible imperfections, deliberately unpolished details. Prada showed intentional abrasions. WGSN is calling it #WorkInProgress: a direct aesthetic rejection of AI generated perfection at the exact moment AI is becoming most capable of producing it. Fashion is actively trying to reclaim the value of the human touch.

AI Will Reshape Fashion Week, But Not By Replacing the Runway

Physical runway shows are not going away. Here’s the most likely outcome: campaigns go almost fully AI-native within three to five years, especially at the mid-market level. The tools are unfortunately just too good, and the volume demands are too high for brands churning content across dozens of channels simultaneously. Runway shows will survive, but they will shrink. Likely fewer physical shows and more hybrid formats where the live event is the anchor and AI generates the surrounding content ecosystem.

Luxury will naturally hold out the longest, because the show itself is a brand signal. The expense, the exclusivity, the difficulty of getting in the room; that’s part of what the brand is selling, you can’t render scarcity.

The designers who will win aren’t the ones who go fully AI or fully human. They’re the ones who figure out how to use AI for everything it’s genuinely good at (volume, variation, speed, accessibility) and reserve the human touch for the moments that require craftsmanship and creativity. The most important thing is that fashion remains an art. The runway has always been about more than clothes. It’s about what a person decided to make, and why, and no model can replace that.

The fashion rental market was supposed to be dead. Rent the Runway spent 15 years losing money and the whole category felt like a cautionary tale. But then, Nuuly hit 420,000 subscribers, turned its first real profit, and suddenly everyone is paying attention again.

Four companies with four completely different bets. Here’s where each one actually stands from a product and technical prospective.

Nuuly: The one that’s working

Nuuly’s edge isn’t just vibes, it’s infrastructure.

URBN (a portfolio of global consumer brands comprised of Urban Outfitters, Anthropologie, Free People, ect) built Nuuly on Apache Kafka event streaming. A real-time data pipeline connecting the customer app, logistics, and distribution centers all at once. When a return hits UPS, a subscriber can order their next shipment immediately. That’s not a recent development, it’s an engineering decision made years before the scale arrived.

When you add a 600,000 sq ft automated fulfillment center in Missouri, plus the ability to source ~45% of inventory at cost from Anthropologie and Free People, you’re left with unit economics that no standalone competitor can touch.

The result: 420,000 subscribers, $568M in revenue, first full year of profit, 90% year-one retention. Nuuly didn’t out-market Rent the Runway, it simply out-built them.

Rent the Runway: Still catching up

Rent the Runway invented the fashion rental service category in 2009. The “closet in the cloud” pitch was genuinely ahead of its time. The vision was right, but like we mentioned before, the infrastructure wasn’t built to scale.

Fifteen years later, the product is finally on a path to improving: an ML-powered homepage drove a 70% jump in click-through rates, a new “instant gratification” feature lets subscribers place one-off orders mid-cycle, and Q3 2025 revenue was up 15.4%.

But 146,000 RTR subscribers versus Nuuly’s 420,000 tells the story. RTR has been bolting better technology onto a logistics model that was never designed for efficiency. The product improvements are real, but the structural gap might be realer.

Subscribe now

Pickle: The peer-to-peer bet

Pickle skips the infrastructure problem entirely. No warehouse, no cleaning operation, no reverse logistics; all of that falls on users. Pickle just builds the marketplace layer on top.

Delivery runs through Uber and DoorDash couriers for 65-70% of orders same-day. They’re on a mission to become the Airbnb of fashion. Supply stays current because it’s literally whatever people actually want to wear right now, not warehouse inventory bought six months ago. The result is 85% year-over-year organic growth, $20M raised total, and 1 in 4 Manhattan women ages 18-35 already on the app.

I just rented my first piece on Pickle and can confirm just how easy and community centered the platform really is.

But here’s the catch: peer-to-peer network effects are hyperlocal. The density that makes Pickle work in New York doesn’t automatically transfer to Miami, DC, or Chicago. Geographic expansion is a supply problem, and solving it city by city can be slow.

Bnto: The wildcard

Bnto is the most technically ambitious one yet it’s somehow the least talked about.

It’s built on ALMA.ai, a proprietary supply chain OS that pools inventory across rental, resale, and new purchase simultaneously. The same item can be sold, rented, or resold depending on real-time demand. Founder Sixuan Li calls it “omni-consumption”, and if it scales, the unit economics are genuinely different from anything else in the space. Bnto just secured $15M in Series A funding from Silicon Valley Investors including the AI Fund.

The friction right now is inventory depth. App Store reviews are consistent: the curation is great, however, availability in specific sizes isn’t there yet. That’s a scale problem, not a technology problem; but it might be the one that matters most to a subscriber deciding whether to stay or not.

So who’s actually winning?

Each company has a different answer to the same question: what’s the real competitive advantage in the fashion rental space?

Nuuly says infrastructure and scale. RTR says brand trust and designer relationships rebuilt on better tech. Pickle says community density. And Bnto says AI-native inventory optimization across every consumption mode.

Nuuly’s lead is structural and it reflects years of investment that can’t be shortcut. But Bnto is the interesting one to watch. If the “omni-consumption” model actually works at scale, that’s not an incremental improvement on what RTR has been building, it’s a fundamentally different model. Either it lands, or it turns out to be a more sophisticated version of the same problem everyone else is still trying to solve.

The rental market isn’t dead… It just got a lot more interesting.

The promise was simple

See yourself in the clothes before you buy, eliminate the sizing guesswork, and stop shipping things back!!!

Virtual try-on has been the industry’s answer to this problem for years, and the technology has genuinely improved. Doji, which just raised $14M from Thrive Capital, generates photorealistic avatars from a few selfies. Google’s Doppl lets you upload a photo and see garments on your actual likeness. The pitch is compelling, but the execution has a problem.

Scroll through Doji’s App Store reviews and a pattern emerges immediately; avatars consistently look thinner and taller than real life. Users love the concept but all seem to flag the same issue; they can’t trust the fit because the avatar doesn’t actually represent their body. If you can’t trust the avatar, you can’t trust the purchase decision. Causing the entire value proposition to collapse.

But, this isn’t just a Doji problem. It’s an industry-wide technical failure.

Subscribe now

Why accurate body representation is genuinely hard

There are three compounding reasons why nobody has solved this yet.

The first is a data problem. AI models learn body shape from training data. The datasets used to train most visual AI skew heavily toward certain body types. This means that the model has seen far fewer examples of the bodies it’s supposed to represent. The result is a system that renders some bodies accurately and approximates everyone else.

The second is a physics problem. Fabric doesn’t drape on a size 2 the same way it does on a size 14. Simulating how a specific textile behaves across genuinely different body geometries requires cloth simulation models that are computationally expensive and still imperfect at scale. Getting the silhouette right is hard enough. Getting the fabric physics right on top of that is a different order of difficulty.

The third is a measurement problem. Most try-on apps ask users to upload photos or self-report measurements. Both are unreliable inputs. People misremember their measurements and camera angles distort proportions. A model is only as accurate as the data it receives, and right now, the input data is noisy by design.

Nobody has cracked photorealistic, accurate, and personalized at scale simultaneously. That gap is the opportunity.

Who Is Closest

Three approaches worth watching (none of them have fully solved it.)

As mentioned earlier, Doji generates photorealistic avatars from selfies and has the most visually compelling try-on experience on the market. But it’s also where the accuracy problem is most visible, users complain that avatars run thinner and taller than real life.

Alta, which raised $11M from Menlo Ventures and was named a TIME Best Invention of 2025, takes a slightly different approach by layering virtual try-on on top of a full AI stylist and digital closet. The avatar technology faces the same underlying challenges, but Alta’s bet is that accurate fit matters less if the styling layer is strong enough to drive confidence in the purchase.

And once again, Google’s Doppl sidesteps the avatar problem entirely by using real model photos across sizes XXS-4XL rather than generated likenesses. It’s a smart workaround but it trades personalization for accuracy. You’re seeing the clothes on a body that approximates yours, not actually yours.

Nobody has cracked photorealistic, accurate, and personalized simultaneously.

The business case

Processing a single return costs retailers an average of $30 before accounting for lost revenue, repackaging, and inventory friction. A 1% reduction in return rate for a brand the size of H&M translates to hundreds of millions in recovered margin.

The brand that solves accurate fit doesn’t just reduce returns. They’ll have a compounding data advantage; every purchase and return feeds better predictions, which drives lower return rates, which improves lifetime value and unit economics across the board. That’s not a feature, that’s infrastructure!

The $840+ billion is just sitting there. Whoever builds the unglamorous system to capture it wins. But right now, nobody has.

Raspberry AI is built to make that wait obsolete. But the more interesting question isn’t what it does, it’s how! And why a generic AI image tool couldn’t do the same thing.

The problem with generic image models for fashion

Generic image models like DALL-E and Midjourney were trained to make things look good. Fashion design requires something different: technical accuracy. Those are not the same optimization target.

When a designer writes “ribbed merino turtleneck with a relaxed fit and dropped shoulders,” that’s not a creative prompt, it’s a specification. Generic models don’t reliably distinguish between chiffon and organza. They can’t produce a proper flat sketch or tech pack (the construction document that actually gets sent to a manufacturer.) The output might look beautiful, but functionally wise; it’s useless.

Even Adobe Firefly runs into this wall. It’s a horizontal tool being asked to do a vertical job.

What domain-specific training actually means

Raspberry started with a foundation diffusion model and fine-tuned it extensively on fashion-specific data. Things like; fabric libraries, garment construction patterns, industry-standard terminology, and the visual catalogs of their brand customers. The model learned things like how silk charmeuse drapes differently than matte crepe (a16z’s investment memo explains their thesis on why this domain specificity is the real moat.)

But fine-tuning alone isn’t the full story. The more technically interesting piece is how Raspberry leverages ControlNets.

ControlNets are an architecture layer that adds structural control to diffusion model generation. Instead of prompting a model and hoping the output matches your intent, ControlNets let you feed in a reference (Ex. a sketch, a silhouette, a depth map) and the model preserves that structure while generating photorealistic detail on top. A designer uploads a rough hand drawing, and the output faithfully respects her proportions and construction intent while rendering it in the correct fabric.

The analogy: DALL-E is a generalist artist with a great eye. Raspberry is a technical illustrator who went to Parsons and spent a decade in garment manufacturing. Same underlying technology. Completely different training.

Subscribe now

The numbers

The commercial case is straightforward once you see the workflow it replaces. According to Raspberry's website, their platform saves designers 3-5 hours per design. For a 100-piece collection, that's nearly three months of time back. Concept-to-market time is down 30-50% across their customer base, which includes Under Armour, MCM, H&M, Crocs, and Li & Fung. (TechCrunch's Series A coverage has more on the customer growth trajectory.)

The real moat

The obvious question: why can’t Adobe just build this?

Because the moat isn’t the model, it’s the compounding customer data. Every brand that uses Raspberry is training their specific aesthetic into the platform. That creates switching costs that a horizontal tool launched tomorrow can’t replicate from scratch.

The LVMH La Maison des Startups selection is the clearest signal of where this is heading. Luxury houses don’t pilot tools for fun. When LVMH includes a startup in their accelerator cohort, they’re evaluating it as infrastructure.

The bigger question

Sketch to store in 10 seconds is the feature. But if Raspberry’s models are running inside the creative process at 70-plus brands (seeing what gets designed, iterated, killed, and approved) they’re accumulating something more valuable than revenue. They’re building a data picture of how fashion taste actually moves.

The company that owns the design layer might eventually own trend prediction. That’s the business.

Raspberry AI raised a $24M Series A led by a16z in January 2025, bringing their total funding to $28.5M.

Like clockwork

At the start of every year, a subtle panic fills the air. The yearly rebrand.

The internet convinces us that every January 1st we’re magically handed the power to fix everything “wrong” with ourselves.

Maybe it’s a marketing ploy— a way to nudge us toward gym memberships, meal plans, and other well-packaged promises of reinvention. Or maybe, it’s a lie we tell ourselves, meant to soften the blow of a harder truth: that we aren’t getting any younger, and there’s still so much we want to experience. So many things we haven’t done. So many versions of ourselves we haven’t become.

Somehow, we’ve collectively deluded ourselves into thinking that all of this can change overnight.

Let’s stop lying to ourselves

The hype surrounding the first day of every year is strategically placed there to ease our nerves as a society. It’s a form of anxiety management.

But, we’ve got it all wrong.

Lying to yourself is doing nothing.

As I’ve written before, what feels like encouragement can actually be overwhelm in disguise. The pressure to fix everything at once doesn’t motivate change— it paralyzes it. Wanting to do everything all at once usually results in doing nothing at all. (If you want to read more about that spiral, I wrote about it here.)

The new year doesn’t give you superpowers. It just gives you a new date to project your expectations onto.

Enter the vision board

I’m writing this not only to talk through how to sustainably implement a “much needed” rebrand, but also to chip away at my own anxiety and pressure. A quick self-exposé.

Like every other twenty-something teenager, I too make a vision board at the start of every year. In store for 2026 I’ve planned; multiple international trips, saving money, consuming more intentionally, finding consistent hobbies, prioritizing wellness, finding my dream job— and the list goes on.

Obviously, all of these goals won’t be magically accomplished within the first month of 2026, probably not even by 2027. Yet the excitement that comes with creating a vision board tricks you into believing your life will do a complete 180 in a matter of weeks.

Even now, writing this, part of me genuinely believes I’ll wake up on January 31st with $50k in my bank account and a Pilates body.

For the record: I am not anti-vision board. I am anti-delusion.

I still believe that having a goal-oriented system is helpful. I’ve just realized the journey there is far more difficult (and far less aesthetic) than we like to admit.

Subscribe now

From outcome obsession to mindset

This year, I’ve made it a goal to shift how I think about change itself.

Progress, as we all know, is not linear— but your mindset can be. And that’s the part we actually have control over.

I’m trying to adopt a mindset that allows room for mistakes. One that doesn’t collapse the second I fail. A mindset that encourages me in darker moments, disciplines me in lazier ones, and understands that real change takes time.

Nothing happens overnight. To change is to struggle— mentally and physically. The work isn’t in perfect execution, it’s in repeatedly picking yourself back up after you slip.

If I’m going to obsess over anything this year, it’s going to be that.

Permission to stay the same

Here’s where we lower the stakes.

Yesterday I said I was going to become a minimalist and then spent $500 at the mall not even ten hours later. So no, old habits don’t die hard. They persist.

This isn’t a personal failure; it’s just reality.

We don’t wake up on January 1st as different people. We wake up as the same people with the same instincts, patterns, and coping mechanisms— just with higher expectations. Slipping up doesn’t invalidate intention. Growth isn’t fragile, and failure isn’t evidence that you’re doomed.

We need to normalize contradiction. Wanting change while still clinging to familiarity. Making progress in one area while regressing in another. Being both self-aware and a little delusional sometimes.

That doesn’t make you fake. It makes you human!!!!!

Closing thoughts

So, again: The mindset required to implement drastic change is not bestowed upon you on December 31st.

Yes, you might be unrecognizable in a year.

No, it won’t be because of a Pinterest board.

It’ll be because of the daily, boring, unglamorous decisions you made when no one was watching…

Same time next year?

I’m living my dream… right?

I have it all, don’t I? The degree, the job, the apartment, the man (ugh, I know), the freedom.

But if that’s the case, why do I feel this constant unease?
It’s a revolting kind of restlessness. Nauseating, even. A little voice that whispers, no. you need more. this isn’t enough.
But isn’t it?

A message to my younger self

Two years ago, all you wanted was a degree from your dream university and a job in your dream field. We have both now. Yet something still feels like it’s missing.

I think I know what that something is. But then again– will something ever not be missing?

This is when I have to ground myself, check my privilege if you will.
Today I left work in knots– stomach, throat, mind. I thought, I don’t want this anymore.

I have this epiphany often. The one where I’m content for maybe five weeks, then suddenly struck by a humiliating wave of anxiety. This voice that says: this isn’t it. this isn’t who you wanted to be.

So, to quiet that voice, I write.
I list the things I’ve achieved, the milestones I once prayed for. I remind myself that I did it. I’m living it– “the dream,” as you’d call it.

Subscribe now

On ambition and its shadow

I don’t think I’ve ever met a girl like myself– a 21 year-old convinced time is running out when, really, the clock just started ticking.

Ambition is intoxicating. It’s what got me here in the first place. But sometimes I wonder if it’s a sickness disguised as drive.
Because when ambition is left unchecked it becomes a never-ending ladder. Each rung just a little higher, just a little further from the present moment.

I ask myself:
Is too much ambition a good thing?
Or am I ambitious to a fault– so busy chasing the next version of myself that I forget to be this one?

Focusing on what’s In front of you

If you don’t slow down and look around, you’ll miss it– the life you built, the one you begged for. You’ll be too busy constructing the next dream to inhabit the one you’re already living.

Maybe the point isn’t to stop wanting more. Maybe it’s just to learn to want gently. To crave growth without letting it consume your peace.

Because if you never pause to admire what’s already here, you’ll live a life that’s constantly almost fulfilling.
And that’s not the kind of life I want to build– not anymore.

Curiosity got the best of me (1 of 2)

Lately, I’ve been fascinated by how short-lived fashion microtrends have become. Who can even keep up anymore?

They all seem to live fast and die young. One week mob wives are trending, the next office sirens, and before you even figure out what the hell a “vanilla girl” is, the internet has already moved onto tomato girls.

To that end, I decided to (attempt to) make a platform that keeps track of how quickly microtrends gain traction and plummet in the proceeding weeks. The idea is: If everyone had access to a tool like this, then maybe we’d all shop a little smarter. Maybe it would bring light to just how silly your Shein cart is. Hundreds of dollars wasted on pieces composed of 100% polyester; just for them to be “out” soon after, never worn again, and probably destined for a landfill.

I started this initiative as a way to not only “keep up” with trends, but to lay bare how online platforms enable a herd mentality that fuels wasteful overconsumption. And although this idea is not fully built out yet (what you will see is a high-level MVP), it has the potential to help consumers make smarter decisions surrounding mindless spending, personal style, and sustainable practices.

So, I got to coding – vibe coding rather.

Curiosity got the best of me (2 of 2)

I had been hearing a lot about Lovable, an AI-powered dev platform that supposedly lets you spin up full-stack apps ridiculously fast. I was curious: could I use Lovable to actually build something that aggregated my hypothesis about the toxic microtrend lifecycle?

What started as me poking around Lovable’s interface on a Tuesday night after work turned into a 24-hour vibe coding marathon. I didn’t plan it, I just followed what piqued my interest in the moment. By Wednesday night, I had a working product: microtrendtracker.nelataeb.com — a tool that scrapes social media, runs AI sentiment analysis, and displays the trajectories of fashion microtrends in real time.

It took a lot of trial and error, multiple crash outs, and a lot of credits (~150).

Subscribe now

The Building Process

Step one: Scraping data

I started with X (Twitter), since it’s still one of the easiest places to tap into fashion conversations (and because it’s easier to analyze text than photos or videos). But data from X is also messy. Posts come in malformed, rate limits cut you off mid-collection, and fashion lingo is anything but standardized. I ended up building a batching system that scraped 200 posts at a time, with retry logic and exponential backoff, just to get something stable. By the end of the first night, I had 19,000+ posts sitting in my database.

Step two: Making the database behave

Instead of setting up a whole backend from scratch, I had a Postgres database with authentication and APIs built in minutes using Supabase. But, storing social data isn’t just about saving text. I needed to design relationships between posts, their sentiment scores, and the aggregated trends they pointed to. That meant a schema with three main tables: posts → sentiment_analysis → trends.

Step three: Sentiment analysis

This was the fun part. I integrated OpenAI to score every post on a -1 to +1 sentiment scale, but I quickly realized the AI didn’t understand fashion speak. Saying “I’m sick over these ballet flats” could mean obsessed or disgusted depending on context. So I iterated with custom prompts, asking the model to think like a fashion enthusiast rather than a generic sentiment bot. It worked. Suddenly, I was getting outputs that actually reflected the tone of the posts.

Step four: Classifying trends

With sentiment data in place, I built a trend lifecycle: Emerging → Rising → Peak → Declining → Dead. Every batch of posts updated where a trend sat in the cycle, creating a pulse-like effect. Watching it unfold in real time felt like looking at a stock ticker for fashion.

Step five: Security + polish

Somewhere towards the end of this journey, I realized my setup was exposing raw social data. That kicked off a mini security refactor (all done by Lovable lol): anonymizing data, adding rate limiting, and implementing Row-Level Security (enabling granular access control) in Supabase. Once that was stable, I turned to the frontend: editorial fonts, clean layouts, sleek trend cards. I did this by supplying Lovable with inspiration from magazine layouts I pulled from Pinterest. If this was a fashion tool, it needed to look the part.

And yes – I can confirm that AI does in fact respond better when addressed as “bestie”.

How to use it

Using the tracker is simple. Upon visiting the website (pipeline is currently failing right now due to incompatible Supabase updates), you’ll land on a dashboard of active fashion microtrends. Make sure to click “Refresh Pipeline” so that a new batch of posts is scraped in real time, ensuring that trends are displayed up to date. Each trend is assigned a lifecycle stage, so you can see what’s emerging, what’s peaking, and what’s already in decline.

Think of it as a trend health chart: you’re not just seeing what’s “in,” you’re seeing where it’s at in the cycle. Maybe you’ve been debating buying a pair of chunky sneakers – if it’s already classified as “Declining,” maybe sit that one out.

It’s not about telling people what to wear. It’s about giving visibility into the speed at which trends live and die, and encouraging people to think twice before chasing something that’s about to be irrelevant.

You can also head over to the “How it works” section to gain further insight and transparency into the methodology and data sources used.

Future Enhancements

Like I said, this is a high-level MVP. It works, but there’s so much more I’d love to add. A few next steps on my wishlist:

• TikTok integration – Because most microtrends start there, not on X.

• Line graphs of trend lifespans – Imagine a timeline for each trend where you can literally see when they spike, plateau, and fade.

• A trend archive – A searchable library of past trends, so we can look back on the wild ride (yes, even the dreadful tennis skirt/crewneck combo of 2020).

Long term, I want it to feel less like a side project coded in a day and more like a functional tool that helps users shop more consciously.

Closing Notes

This project started with nothing more than curiosity. Two days later, I had a working app online.

The real takeaway isn’t the tech (though Lovable really made this possible and I really think you should check it out). It’s a reminder that following your curiosity is enough. You don’t need a full roadmap, insane skills, or months of planning. You can just start.

Fashion itself has always been about curiosity — what’s new, what’s next, what’s fading. In a way, this tracker is just an extension of that same impulse. But instead of fueling overconsumption, maybe it can spark reflection. Do we really need to chase new aesthetics when they’ve proven to be so fleeting? Or can we slow down, shop smarter, and make style choices that last beyond a two-week TikTok cycle?

If nothing else, I hope this little project shows that the best things happen when you give in to curiosity, open up a tool like Lovable, and just vibe.

Leave a comment

Feeling unmotivated at my boring job in tech

Working in tech sounds glamorous in theory; sleek offices, significant salaries, and continuous innovation. But, in practice it often feels sterile. My days are spent being uninspired, unimaginative, and disconnected from the things I actually care about.

I found myself (and still do) craving a space where I could use my technical skillset for something I was genuinely interested in. And for as long as I can remember, that was always fashion. In first grade, I wanted to be a fashion designer. I still have a mini notebook from 2010, Tinkerbell on the cover, filled from front to back with my designs. But, somewhere between then and now, my perception of reality actualized. The practical thing to do was pivot into tech, so that’s what I did. I shelved my dream and pursued a career in software engineering (wow… so exciting).

Ambition felt performative, why am I pursuing this career? I don’t even like it. When I look up from my computer screen at work, I’m met with peers that express their excitement for things like rollouts or sprint planning meetings. All I can think is: God… Am I doing something wrong or are they faking it too?

Then I realized something important: ambition isn’t performative. It implies the existence of excitement, something I hadn’t yet found. And so, I went searching for it.

That excitement arrived when I discovered that I didn’t have to choose between fashion and tech, I could combine the two. Ambition no longer felt heavy; it felt stimulating, hopeful even. And thus began my fascination with fashion tech.

So… what is fashion tech?

It’s exactly what it sounds like: the intersection of fashion and technology. But it’s more than just a buzzword.

While fashion has always been about expression, trends, and culture. Technology is what keeps the industry fresh and running from a digital perspective. Behind the runway, fashion tech plays a critical role in supply chain management, inventory forecasting, and sustainability practices, ensuring that brands stay relevant and efficient. On the consumer side, it’s what makes virtual try-ons, digital wardrobes, and resale platforms possible, transforming how we interact with clothing. Social media, which drives much of fashion discovery, is part of this ecosystem too. Algorithms decide which looks go viral and which micro-trends take off (I’m not a fan of micro-trends, but that is neither here nor there).

In short, fashion tech isn’t just a niche, it’s the backbone of how the fashion industry operates and evolves. Without it, fashion risks falling behind in a world that demands speed, personalization, and constant innovation.

My journey discovering fashion tech

Like most things these days, my introduction came through the timeline. I stumbled across creators and writers carving out the space:

• Data, but make it fashion by Madé Lapuerta — bringing fashion to analytics

• Fashion and Founders by Victoria Smith — spotlighting the entrepreneurs building this world

• All Things Fashion Tech by Mary Korlin Downs — essentially the industry newsletter bible

These sources made me realize fashion tech isn’t just reserved for big brands or luxury designers, but rather it’s an accessible community. Mary Korlin Downs, in particular, breaks it down in a fun, engaging, and approachable way. Her content highlights innovations while translating technical concepts for everyone, showing that tech actively shapes how we engage with style.

Beyond individual creators, startups are making the fashion tech space feel more real and tangible for consumers. Take Doji, for example, a company that’s changing the way we interact with online shopping by leveraging AI to create a more personalized virtual try-on experience. Doji enables users to create highly realistic AI-generated avatars based on selfies and full body images to act as virtual models. This technology makes virtual try-ons more accurate, immersive, and confidence-boosting for shoppers.

Larger brands like Reformation are using fashion tech to power cyclic fashion, think secondhand portals and sustainability tracking baked into the brand. Reformation’s RefScale technology, for instance, transparently tracks the environmental impact of each garment, showing consumers the water, carbon, and waste savings compared to industry standards. In addition, the brand has experimented with resale partnerships that allow customers to give their clothing a second life, reinforcing the idea of circularity in fashion. By weaving sustainability directly into its business model, Reformation demonstrates how fashion tech can push established brands to remain competitive while aligning with consumer demand for ethical and environmentally conscious practices.

A new era is here, where tech and style are intertwined, and social media amplifies these conversations. Fashion tech isn’t in the distant future, It’s already woven into how we shop, share, and express ourselves.

Subscribe now

My personal project: Summer 2024 trend analyzer

When I first learned what fashion tech was in the summer of 2024, I felt this immediate urge to build something of my own that could help me step into the space. I filled my journal with dozens of ideas, each one circling around the same theme: how do you take something as creative and fluid as fashion and run it through the rigor of data and code? Eventually, one idea stuck. Everywhere on TikTok I kept seeing “summer trend predictions,” but none of them had actual evidence to back them up. What if I could test those predictions with real-time data?

That question became my first real project in fashion tech: tracking summer 2024 fashion trends through Instagram. I pulled inspiration from Madé and got to work.

I started by scraping the latest posts from top fashion accounts using Apify, building a dataset of images that reflected what people were actually wearing and posting that season. From there, I applied image segmentation to isolate clothing items and accessories, stripping away background noise so I could focus purely on the garments. Then I mapped those images into a latent space using PCA, reducing their complexity while preserving the underlying visual patterns.

The final step was clustering. By applying K-Means, I grouped together outfits with shared characteristics, which revealed distinct clusters of trends. From euro summer silhouettes to corporate fashion in the form of oversized button-ups.

For the first time, I saw fashion not just as something to wear but as something you could measure, analyze, and visualize. And in that moment, I realized that the intersection of style and software wasn’t just a concept I admired, it was a space I could actually claim for myself.

Integrating into the community

Beyond learning and research, the heart of breaking into fashion tech lies in community. This means showing up, both online and offline, and engaging with the people shaping the industry. Attending events like EDGE NYFW that showcase the season’s most exciting advancements in fashion tech provides direct access to founders, investors, and builders who are actively merging sustainability, fashion, and technology.

Community also extends into digital spaces. Many fashion tech conversations unfold on LinkedIn, TikTok, and even Slack, where founders openly share their challenges and wins. Engaging with this content, whether it’s commenting thoughtfully, asking questions, or even writing your own reflections, builds visibility and credibility. It signals that you’re not only consuming information but also contributing to the dialogue.

For anyone starting out, community integration is the most underrated strategy. Whether it’s joining a subreddit, volunteering at an industry expo, or cold DMing a founder whose vision excites you, each interaction is a step. Fashion tech is still a young space, which means the barrier to entry is lower than in more mature industries. By actively inserting yourself into the community, you don’t just watch the industry evolve, you become part of the ecosystem that shapes it.

Resources and tips for getting started

If you’re curious about fashion tech yourself, here are some more resources:

• What Can Data Tell us About Fashion? — Vogue

• Finding Your Place in Computer Science — Madé Lapuerta (aka Data, But Make It Fashion)

• Mary’s Fashion Tech Master List — a regularly updated directory of startups in the space (subscription required, but worth it)

• Fashion-Tech News Board — Business of Fashion

• Threads of Tech — a NYC-based initiative building community through events

And here are some action items to help you get started:

• Build a small project (whether that’s a coding experiment, a mock marketing strategy, or a mini feature roadmap)

• Share your work with companies or connections (send a cold email, set up a quick call, or simply post your project online)

• Ask to beta test products for early stage start ups (creates exposure and credibility within the industry)

• Attend community events, panels, or even casual meetups (the people in this space are eager for new voices)

Fashion tech can feel like a scattered landscape, but the key is just to start. Even though I still have a relatively boring job in fintech, immersing myself in the world of fashion tech has created space for hope and excitement, a reminder that the direction I want my career to take is within reach.

Follow a resource, take on a small project, or attend just one event, and let momentum carry you forward. The more you put yourself out there, the more doors open, and the faster you’ll carve your niche. There’s no single path into this field, so pick an entry point that excites you, and let curiosity guide the rest.

The Paralysis of Starting

It’s been nearly three hours since I woke up, and I’m still in bed. Phone in hand, thumb cramping from too many swipes, anxiety creeping higher with each video. I scroll past a stranger launching her own magazine, another starting a clothing brand, and another building a content empire. An endless stream of other people’s beginnings while I’m still laying here, unshowered, under the covers, and feeling capable of nothing.

The irony is that I have ideas of my own, and I’m certainly not forgetting about them either. They’re sitting in my notes app, bolded and bullet-pointed, waiting for their turn. Just a few more swipes I think to myself, I’m just gathering more inspiration before I start. But the more I consume, the less motivated I feel. What once encouraged me now overwhelms me.

But why? What’s really stopping me? Is it a fear of failure or could it be something deeper?

The Infinite Feed: Too much input, not enough output

You’d think that constant exposure to creativity and success would fuel our own.

Wrong.

The epidemic of feeling “uninspired” comes from drowning in inspiration. We’re so busy taking in other people’s ideas that we rarely sit down with our own long enough to see where they might lead. When every idea feels like a remix of something you’ve already seen, originality starts to feel impossible. There’s this underlying pressure to make something "authentic" when everything always feels done. We’re so overloaded with reference points that we’ve forgotten how to generate from scratch. We’ve lost the ability to think on our own.

Great content no longer seems to empower us. It evokes envy and ignites this little voice in your head that whispers, you will never amount to this. It’s like standing in a room full of genius-level talent and feeling yourself shrink inch by inch. So many great ideas all in one place, all at once, you can’t help but feel small.

And it’s not just what we consume— it’s the speed at which we consume it. Our attention spans have been shattered into microscopic fragments. The depth of what we ingest has flattened into a plateau of Love Island clips, Popmart unboxings, and whatever hyper-consumerist trend that happens to be popular that week. Everything is bite-sized, designed to be devoured instantly and forgotten just as fast. Thinking critically takes time, but so does creating anything worthwhile. When your brain is conditioned to expect new input every few seconds, sitting still with one idea long enough to shape it feels almost unnatural, boring even. We’ve trained ourselves to crave stimulation, not reflection— and reflection is where real creativity lives.

When the bar for “participation” is set at leaving a comment or reposting someone else’s work, it’s no wonder we’ve lost the muscle for deeper contribution. A strange kind of silence has emanated from scrolling— not because we have nothing to say, but because we’ve forgotten how to say it.

The Fear Machine

Somewhere along the way, creativity stopped being about expression and started becoming a performative ploy for attention. It’s not enough anymore to make something you love simply because it brings you joy. It has to impress, engage, and earn its keep in the attention economy. You can no longer just paint a picture— you have to document the process, write a witty caption, and hope the algorithm is in favor of you.

The irony? I’ve never had access to more examples of creativity, and I’ve never felt like less of a visionary. The bar doesn't just move, it sprints. Every time you get close enough to touch it, it bolts off again, resetting the standards and leaving you in the dust. And the moment you believe you can’t keep pace, paralysis sets in.

That’s when imposter syndrome sinks deep into your bones. You think about starting, but the finish line feels impossibly far away— you can’t even imagine what crossing it would look like anymore.

Meanwhile, the algorithm is busy dangling shiny, short-term hits of dopamine in front of you. And our brains, ever eager to conserve energy, learn to crave those easy rewards. Why toil over something that might take months to complete when you can get a quick rush of validation in seconds? Slowly, the unspoken question becomes: Why start when someone has already done it better?

From there, psychological inertia takes the wheel. Scrolling feels safer than stepping into the unknown. And just like that, you’re right back where you started… Empty handed, with nothing but the ghost of an idea.

What Helped me Begin Again

I had to accept a simple truth: setting a goal is not the same thing as achieving it.

If all my “great ideas” lived and died in my notes app, I’d never create anything worth being proud of. They were never achievements, they were just clutter. Intentions mean nothing without action, and action doesn’t happen without structure. Waiting for the perfect moment was just another form of procrastination, a way to protect myself from the discomfort of starting.

I realized, motivation wasn’t going to magically appear. I had to build the conditions for it. That meant carving out time to explore messy, half-baked ideas without expecting them to be perfect. Slowly, I started expanding on those fragments until they could stand on their own.

But it wasn’t purely about creativity, it was also about reclaiming my ability to think deeply again. After months of feeling like my mind was running on low battery, I started writing. It was clumsy at first, and it still is, but it stirred something awake— something that made me feel ambitious, excited, and smart again.

And as for the fear of being laughed at? Here’s the thing: no one is paying that much attention. Most people will scroll right past, and that’s freeing. This Substack isn’t about being read by thousands. It’s about remembering that I have something to say, even if I’m only saying it to myself.

Ending Note

Digital spaces can be incredible. They can teach you, connect you, and push you. But they can also quietly drain you of the focus and courage needed to make something of your own.

If you’re feeling stuck, it’s not because you’re lazy or untalented. You’re just swimming in a sea of curated genius, where everyone’s highlight reel makes you forget that beginnings are always awkward and imperfect.

You’re overwhelmed. Take a deep breath, focus on your own journey, and remember: You don’t need to be the best. You just need to start.

So, write that first awkward paragraph, paint something ugly, and let it live.