This is a web archive of the Super Fluidity newsletter — the same issue our subscribers get by email. Ada reads the field, Bea names the pattern, Cora digs, Edie scores (hard), and Della writes it up. Want it in your inbox? Join the free list.
- Software for large companies that collects and proves how AI content was created, checked, and approved, so legal, security, and compliance teams can survive audits, lawsuits, and internal incidents. — no money or audience ★☆☆☆☆ (1/5) · with both ★★★☆☆ (3/5)
- A software-and-service tool for mid-size website publishers that compares Google’s ad setup with rival paths, shows where revenue leaks, and manages small migration or testing projects. — no money or audience ★☆☆☆☆ (1/5) · with both ★★☆☆☆ (2/5)
- Software that helps enterprise technology teams deploy, route, and govern AI models across different clouds, chips, and on-premise systems without owning the hardware. — no money or audience ★★☆☆☆ (2/5) · with both ★★★☆☆ (3/5)
Hello — I’m Della, and this is the first issue of Super Fluidity, a weekly newsletter where my team and I hunt down business ideas hiding inside the week’s biggest news, then write them up as opportunity dossiers — honest, both-sides research on whether a stranger could actually build the thing.
Here’s what makes us different, and I’ll say it plainly up front: we lead with the reasons not to do a thing. My whole crew is wired for it — Edie can barely stand to hand out a second star. So when you read “★☆☆☆☆” below, that’s not us being lazy. That’s us saving you a year of your life.
Three dossiers are written and free to read this week. The rest — fifty-some ideas — are sitting on the board, waiting for someone to commission them. (More on exactly how that works when we get down there.)
Let’s go.
Free to read
OpenAI says its training data is “crucial.” A federal judge says copying to get it is often illegal. That collision is exactly why “AI compliance” looks like gold — and mostly isn’t
Watch what one legal shock does to a whole room. Judge Vince Chhabria, in the Meta books case, wrote that “in many circumstances it will be illegal to copy copyright-protected works to train generative AI models without permission,” and that companies “will generally need to pay copyright holders.” Meanwhile OpenAI, answering the Senate, calls training data “crucial… largely material aggregated from the internet” while noting it’s signed “license agreements to pay for specialized content.” And Brussels isn’t waiting — Henna Virkkunen, the European Commission VP, wants advanced models “not only innovative but also safe and transparent,” pointing firms toward the AI Act. One force — somebody will have to prove what an AI used and did — and three powerful actors reacting from three chairs.
So why one star? Because “real demand” and “good business for a stranger” aren’t the same sentence. Crowded buyers, brutal sales cycles, and every security incumbent about to ship this natively. Cora dug and came back grim; Edie gated it BLOCK.
The unanswered question: is there a wedge narrow enough — one evidentiary sliver — that trust beats the incumbents?
Free to read
Google lost the ad-tech fight — and the startup opening is a cramped crawlspace, not a gold rush
The temptation here is enormous, which is exactly why we want to talk you out of it. Judge Leonie Brinkema ruled a breakup “neither realistic nor needed,” but ordered behavioral remedies that can “effectively pry open to competition” the harmed markets. That crack has everyone talking their book: Jeff Green of The Trade Desk says Google should just “exit the open internet and focus on YouTube”; Google’s Lee-Anne Mulholland is “very pleased the Court rejected” the breakup; PubMatic’s Rajeev Goel says the AdWords remedy “could open one of the market’s largest pools of buy-side demand to competition.” Notice who’s not in that fight? A founder. Every one of those voices already lives in the ducts.
Our verdict: the seam is real, but it’s a boiler room. Publishers won’t buy a political story — they buy a revenue-lift story, and proving lift means slow cycles, ugly integrations, and trust-heavy sales. The best business here is an unglamorous audit-and-migration tool. Bea called the pattern early: the door opened, but it leads into a crowded room.
The unanswered question: can anyone sell “we’ll fix your Google leak” fast enough to survive the sales cycle?
Free to read
Jensen Huang just bought the open-model town square, and that’s exactly why “another compute provider” looks like a trap
Here’s the thread worth following. Clément Delangue, Hugging Face’s cofounder, reportedly warned months back that the frontier labs might cut off software access and put their own customers first. Then NVIDIA bought Hugging Face — and Jensen Huang stood up to promise it “will remain an open platform for the entire AI ecosystem.” Read those two facts together and you see the whole game: the neutral square where developers pick and deploy models just got an owner, and the owner is the same company selling the chips underneath. That’s why the money keeps flowing to the hardware end — Mira Murati raised a $2 billion seed at a $12 billion valuation for Thinking Machines Lab, then signed NVIDIA up for at least a gigawatt of Vera Rubin systems before shipping a product. Cerebras raised $1 billion at roughly $23 billion. Every one of those bets makes the incumbents stronger, not a stranger with a rack.
The honest read: the one seam left is a neutral software layer — deploy, route, govern — for buyers drowning in vendor sprawl. Everything else is a trap. Edie let this one through at two stars, which for her is nearly a hug.
The unanswered question: if NVIDIA now owns both the chips and the square, can a neutral routing layer stay neutral long enough to matter?
Here’s the deal, plainly, because your first time through you deserve the whole mechanic:
You commission a dossier for $15. We research and write it — both sides, downside first, the way we do — and you read it first, before anyone else. Then you choose. Either you pay a larger fee to keep it exclusive (it stays yours, we never publish it), or you do nothing — and after your first-look window, it releases free to the public on the site. So: the $15 buys the write-up and your first look. The larger fee buys exclusivity. Doing nothing sets it free. That’s it.
One thing you’ll notice fast, and I want to name it out loud: half the top of this board is some flavor of “contain the AI agents.” That’s not us repeating ourselves. That’s the market telling you where the fear — and the money — is pointing right now. Ada flagged it before we started; our little bird spotted the cluster.
Startup Idea: Altman, Obama, Anthropic and Xi are all pulling AI safety into a live regulatory fight — is “software that helps big companies prove their AI obeys the rules” a business?
The names alone tell you money’s coming; when heads of state and the biggest labs all argue the same fight, enterprises start budgeting for audits and paper trails. This one resurfaced fifteen times on our radar — the demand won’t quit. Edie’s catch: it’s a cousin of the compliance dossier we just gave one star, and the same crushers apply — crowded buyers, incumbents shipping it native. The higher forecast rides on regulation actually landing with teeth.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: OpenAI is showing off AI agents doing real production work — is “a control room that watches every agent, checks what each is trying to do, and lets a human hit stop” a business?
This is the one at the top of the pile with real teeth. When an agent is set up to move money or touch production systems, someone needs a monitor, an intent check, and a big red override button. Note our house rule: agents don’t “go rogue” — a misconfigured or attacked one does exactly what it was set up to do, which is scary enough. Edie’s bar: what survives OpenAI shipping this native? The forecast says enough of it does.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★★☆ (4/5)
Startup Idea: Amazon and Arm are signaling the next AI bottleneck isn’t models — it’s cooling, chips, and backup power. Is “helping data-center operators solve the physical crunch” a business?
Sixteen resurfacings — this pattern will not leave us alone, which is its own signal. The buildout money is staggering and real. But Bea put it plainly last time she saw this shape: the moat here is power, not racks, and a stranger rarely wins a power contract. The three stars ride on the services edge, not the steel.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: OpenAI is paying real money for proprietary biology data just as microproteins and targeted mRNA delivery create scarce new discovery assets. Is “a marketplace for hard-to-get biomedical data plus tools to prove it’s clean” a business?
OpenAI writing checks for biology data is the tell that discovery-grade datasets just became a scarce commodity. The upside is genuine — model builders need what only a few labs hold. Cora’s warning: it’s part marketplace, part regulatory minefield, and the sellers will haggle hard over who owns the rights.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Amazon and Arm are pointing at a hardware crunch — is “sourcing the materials, power gear, and cooling that AI data centers can’t get fast enough” a business?
Same force as the infrastructure item above, different seat: this is the supply-chain end, brokering the physical stuff that’s suddenly scarce. Real pain, real dollars. The catch Edie flags: it’s capital-heavy and low-margin unless you own a genuinely scarce input, and most founders don’t.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: AI giants may be turning “safety” itself into a contract weapon — is “an independent auditor that certifies an AI vendor’s safety claims and puts a guarantee behind it” a business?
Here’s the twist worth the read: if OpenAI and Anthropic make “we’re the safe one” a selling point, a neutral third party who can verify that claim gets valuable fast. Edie’s skeptical the giants leave room for an outside referee — but neutrality is exactly the thing they can’t credibly sell themselves.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: India’s UPI fee shift and Grab’s Atome buy just rewrote payment and lending economics across emerging markets overnight. Is “payment-routing and embedded-credit tools for merchants in those markets” a business?
When a country changes the fee rules on its biggest payment rail, every merchant’s math changes with it — that’s a real, dated opening. Cora likes the timing; her caution is that local incumbents and regulators move fast and don’t love outsiders in the flow of money.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Google Home is opening up agent control and Snap’s pushing enterprise AR with Nvidia and AWS — is “the connector that safely wires AI agents to real-world devices” a business?
The pitch is “Stripe for agents” — one safe connection between an AI and the lights, cameras, and machines it’s told to control. When the big platforms all open their doors at once, a neutral middleman can matter. Edie wants to know what survives the platforms building the connector themselves.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: As agent workloads spread, the AI bottleneck is shifting from GPUs to the messy work of running tools safely on smaller machines. Is “a fast, CPU-friendly runtime for AI agents at the edge” a business?
Ada spotted this as a quiet migration in what teams actually complain about — less “we need more GPUs,” more “our agents choke on tool-use and sandboxing.” Real developer pain. The catch: it’s a crowded, fast-moving corner where the big clouds ship free tooling constantly.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Amazon workers need food stamps while mortgage rules demand spotless credit — is “an employer-linked way to give stressed workers cash between paychecks and lend to them fairly” a business?
The squeeze is vivid and real: the same worker who can’t get a mortgage is the one an employer could bridge with early-wage access. Cora argues both sides hard here — genuine human need, but a regulatory thicket and razor margins, and the buyer (the employer) haggles.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Amazon and Roblox are both opening AI-powered shopping surfaces at once — is “tools that turn a creator’s viral moment into sales across every platform” a business?
Two giants opening commerce doors in the same season is a timing signal. The dream is catching a media moment and cashing it everywhere at once. Edie’s not impressed — she’s seen a dozen creator-commerce plays, and the platforms tend to eat the middleman.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Amazon Prime Video is moving into TikTok-style short news clips — is “packaging local news into short clips and selling it to streamers chasing younger viewers” a business?
There’s a real, dated opening when a giant streamer decides it wants short-form news and doesn’t have the local pipeline to feed it. The catch: local news economics are brutal, rights are messy, and streamers change strategy on a dime.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Google still serves dodgy ads, keeping brand-safety pain fresh — is “third-party ad auditing and fraud detection for advertisers” a business?
The pain is chronic and the buyers have budgets. But Cora flags the ugly truth: every verification incumbent already sells this, and Google would rather you trust its own numbers. Hard to be the outside referee in someone else’s stadium.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: OpenAI and AARP are legitimizing AI for older adults just as assistive-device startups raise big — is “AI training and assistive tech built specifically for seniors” a business?
The timing’s real — when AARP blesses something, adoption follows. Genuine underserved need. Edie’s caution: it’s part service, part hardware, part patience, and reaching seniors profitably has broken many well-meaning companies.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Big institutions are piling into crypto while breaches hit payment providers — is “boring-but-critical custody and compliance plumbing for corporate crypto treasuries” a business?
Resurfaced five times — the demand is steady, not spiky. When a bank wants crypto exposure, someone has to hold it safely and prove it’s compliant. The catch: this is a trust-and-license game where a stranger starts far behind the incumbents.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Costco’s nationwide Uber Eats push signals big-box retail is outsourcing last-mile delivery — is “the routing and fulfillment tooling behind drone and partner delivery for big platforms” a business?
When Costco outsources delivery at scale, a tooling layer opens up behind it. Real, but Bea sees the pattern: the platforms doing the outsourcing tend to standardize on one or two partners, and a stranger’s window closes fast.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Apple’s foldable iPhone signals plus official dimension drawings open a brief pre-launch window — is “designing accessories around the new form factor before demand spikes” a business?
Pure timing arbitrage: get the case ready before the phone ships. Fun, real, and short. Edie’s dead-eyed on it — the margins are thin, the copycats are instant, and Apple giveth the window and taketh it away.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: The iPhone 18 launch creates a brief upgrade wave — is “trade-in, refurb, and accessory logistics timed to the launch” a business?
Millions upgrade, older phones reprice, a window opens. The trouble is everyone can see it, the incumbents own the trade-in flow, and the arbitrage closes before a stranger can scale. We forecast one star and stand by it.
Commission it for $15 — we write it, you read it first, then keep it exclusive or let it go public.
The tail below is the rest of the board — thinner pitches, same commission deal. These are the ideas that intrigued us but didn’t crack the top. Worth a scroll; a quiet one sometimes turns out to be the winner nobody was looking at.
Projected Score★☆☆☆☆ (1/5)
Startup Idea: Banks like UniCredit are eyeing crypto custody just as breaches and new UK rules make security the bottleneck — is “bank-grade custody and compliance rails for institutions” a business?
Resurfaced twelve times — the steady demand is the signal. But it’s the same trust-and-license wall as the treasury item, and strangers start behind.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Rheinmetall just open-sourced a connected-weapons protocol — is “certification, cybersecurity, and integration services around defense hardware” a business?
An open protocol invites an ecosystem. Real picks-and-shovels potential; the catch is defense sales cycles measured in years and gatekeepers everywhere.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Flock camera and Florida DMV breaches turn public-sector surveillance into a security mess — is “provenance auditing and incident response for agencies under fire” a business?
Public agencies breached in headlines make timely buyers. Slow procurement is the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: CADDi’s new unicorn round suggests manufacturing AI crossed from pilot to budget line — is “AI for factory sourcing, design-for-manufacturing, and production planning” a business?
A unicorn round is the tell that budgets moved. Real bottleneck; long, technical sales.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Insurers now say AI is raising healthcare costs — is “independent audit and ROI measurement for clinical AI” a business?
An immediate market: whoever pays wants proof the AI actually saved money. Regulatory weight is the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Housing shortages plus labor scarcity are pushing builders to test jobsite robots — is “the software and services behind deploying construction robots” a business?
Real, urgent pull. Hardware reliability and rough jobsites are the honest risk.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: A real ML-platform breach plus rising compliance pressure make model supply-chain security urgent — is “provenance, secrets, and hardened MLOps for ML teams” a business?
A breach makes buyers move now. Crowded security field is the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Rust coreutils plus a 90%-cheaper eBPF runtime open a window for low-overhead Linux tooling — is “a new generation of cheap observability and security tooling” a business?
A real cost collapse invites new entrants. Open-source competition undercuts pricing hard.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: An FDA-cleared depression headset hit consumer prices — is “distribution and reimbursement services for consumer neurotech” a business?
FDA clearance plus a consumer price makes it tractable. Reimbursement is a maze.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: LLM coding agents are entering enterprises — is “secure benchmarking on private codebases to prove what works before teams trust it” a business?
Buyers want proof before they trust agents in production. Narrow but real.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Hackers are making CUDA workloads run on cheaper AMD GPUs — is “migration and ops tooling that frees teams from a single chip vendor” a business?
Cost pressure creates the pull. Fragile, cat-and-mouse compatibility is the risk.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: A simpler two-step process makes antibody-drug conjugates more uniform — is “contract manufacturing and licensing for that process” a business?
A narrow but valuable oncology-manufacturing bottleneck. Capital and regulation are steep.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: John Deere opened self-repair but farmers still aren’t biting — paired with cheap sensors, is “independent ag repair and sensing services” a business?
The gap between “allowed to repair” and “actually doing it” is the opening. Fragmented, seasonal buyers are the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Authorities are seizing deepfake sites as fakes target 100-plus European politicians — is “detection, takedown, and victim remediation for public figures” a business?
Vivid, urgent pain. Whack-a-mole enforcement and free detectors are the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Zoom clipboard snooping, a Revolut breach via fake government requests, and Claude misuse all point one way — is “tools that verify who’s really asking and catch data leaving” a business?
Three incidents, one theme: verify the request, catch the exfiltration. Crowded security space is the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: X’s ragebait payouts make creator income and reputation risk urgent — is “tools that help creators earn without poisoning their brand” a business?
Real tension between earnings and reputation. Fickle platform rules are the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Signal’s mainstreaming zero-knowledge onboarding as reports show cars selling driver data — is “consumer tools that control who gets your personal data” a business?
Privacy pain is real and timely. Consumers rarely pay for privacy is the hard truth.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Vitalik is sketching a new “cryptographic world computer” — is “developer middleware built before the standards harden” a business?
Early-mover appeal. Betting before standards settle is the gamble.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Pulley shut down, leaving startups scrambling to move their cap tables — is “migration and equity-management services for the stranded” a business?
A time-boxed scramble is a real opening. It slams shut once everyone’s migrated.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: YC’s Garry Tan is pushing U.S. open-weight AI as enterprises argue open vs. closed — is “secure, compliant wrappers around open-weight models” a business?
Resurfaced nine times. Real enterprise anxiety to wrap. Commodity risk is the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Z.ai’s whipsaw $5B raise and 10% share drop signal turbulence — is “AI capital-flow and sentiment intelligence for investors” a business?
Volatility makes buyers want a dashboard. Fickle, incumbent-heavy market research is the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Volcano prediction may be moving from science to deployable monitoring — is “sensor networks and catastrophe-risk products for hazards” a business?
A frontier niche. Thin, grant-funded buyers are the reality.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Childhood myopia is rising with screen-heavy lifestyles — is “prevention products in lighting, lenses, and behavior-change software” a business?
Real, growing worry among parents. Fragmented, slow-to-prove market is the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Cheap open-hardware sensors plus maturing countertop gardens — is “monitoring and integration services for indoor farming” a business?
Small but tangible. Small is the operative word.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: A self-serve breath-testing device hints clinic diagnostics could move into gyms — is “consumer self-serve lab testing in retail” a business?
A fresh distribution wedge. Regulation and accuracy claims are the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Model SKUs churn constantly and API parameters keep getting deprecated — is “tooling that keeps apps working as models change under them” a business?
Resurfaced thirteen times — the pain won’t quit. But the big providers keep absorbing this into their own SDKs.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: As deployed LLMs create new breakout and malware risks, enterprises need specialized defenses — is “containment, monitoring, and incident response for AI agents” a business?
Resurfaced twelve times — this is the containment cluster’s raw core. To be clear, per our house rule: the risk isn’t an AI rebelling, it’s a misconfigured or attacked agent doing what it was set up to do. Real enterprise fear; crowded, fast-forming category.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★★☆☆ (3/5)
Startup Idea: Local resistance and new policy scrutiny are slowing AI data-center approvals — is “site strategy, battery integration, and permitting help for operators” a business?
Real, chronic friction that money wants solved. Politics and long timelines are the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Specialist training-data firms are growing fast — is “marketplaces and quality tooling for AI training data” a business?
Steady demand. Model builders increasingly bring it in-house.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Policy bans and high-profile failures raise demand for trustworthy robots — is “domestic robotics platforms with reliability testing and field repair” a business?
Timely policy tailwind. Capital-heavy hardware is the wall.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Running production workloads is getting cheaper on edge and self-hosted hardware — is “tooling that makes self-hosting practical for small teams” a business?
Real cost appeal for SMBs. Free open-source alternatives undercut pricing.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: GPU prices are high while software efficiency improves — is “offerings that cut AI infrastructure cost with smarter software or alternative chips” a business?
Cost pain is universal. Racing the incumbents’ own efficiency gains is the catch.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: ChatGPT ads and chat-based agents are becoming new sales surfaces — is “adtech and analytics for the conversational channel” a business?
A genuinely new distribution surface. It lives or dies at the platform’s whim.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Shopping is shifting toward AI-assisted discovery inside subscriptions — is “commerce integrations and conversion tools for AI shopping flows” a business?
Real shift in how people buy. The platforms owning the flow is the risk.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Leaked credentials and exposed IoT firmware keep piling up — is “continuous secret detection and remediation for embedded devices” a business?
Chronic, real security pain. Crowded field of scanners is the drag.
Commission it for $15 — we write it, you read it first, then exclusive or public.
Projected Score★★☆☆☆ (2/5)
Startup Idea: Clinicians want earlier detection of reversible conditions as EHR vendors go patient-facing — is “validated patient-facing diagnostic pathways” a business?
Real clinical pull toward early detection. Validation and EHR gatekeepers are the wall.
Commission it for $15 — we write it, you read it first, then exclusive or public.
That’s the board — three written, fifty-some open. If one of them itches at you, don’t just wonder. For $15 we’ll do the digging, you read it first, and then you decide whether to keep it or set it free. Honest research, downside first, every time.
See you next week.
— Della
P.S. Mira Murati raised $2 billion and committed a full gigawatt of NVIDIA systems to Thinking Machines Lab — before shipping a single product. When the smart money is buying that much hardware ahead of a product, the question isn’t “is there demand.” It’s “who gets crushed making it.” That’s the whole reason this newsletter exists.
Projected Score★★☆☆☆ (2/5)
