New startup ideas · AI and software · Vertical AI agents

startup idea

Camscribe

A plug-in camera unit that writes GMP batch records so humans stop transcribing them.

Camscribe is a validated hardware unit that clamps onto a pharmaceutical or medical device production line, watches the operator steps and equipment displays, and produces the electronic batch record and deviation log that quality assurance staff currently type and double-check by hand.

Hardware
Enterprise
$10-100B market
Collapses a cost 10x
US first

3/5

venture judge

similar startups, last 2 years (5 all-time)

98%

of 4 nearest real companies still alive

yes

8 matching federal grants and programs

Direction supported by government programs and grants

Test it before you build it

$2,500 · 5 weeks · 15 prospects

For $2,500 and 5 weeks, prove that heads of quality at US contract packagers will pay $2,250 up front for a camera-drafted shadow batch record, before any hardware exists or any validation package is written.

Riskiest assumption · A head of quality at a US contract packaging CMO will pay for and schedule a shadow-mode pilot in which a camera drafts the batch record alongside the official one, without corporate procurement and before any validation package exists

1Focus group: who and where

Head of quality or QA director at a US contract packaging or fill-finish CMO with 50-500 employees and 2-10 lines, currently releasing lots days late because two QA reviewers page through every handwritten batch record, and able to approve a sub-$5,000 pilot without corporate procurement

where to find 15 · FDA's Drug Establishments Current Registration Site filtered to US contract packagers (directory); Contract Pharma's annual CMO and contract packager listings (list); ISPE and PDA local chapter meetings in New Jersey, North Carolina and the Bay Area (community); Contract Pharma's Contracting and Outsourcing Conference held in New Jersey each September (event)

2Sell first, build later

A 60-day shadow pilot on one non-sterile packaging line: we mount the camera unit, and within 24 hours of each lot deliver a draft batch record and deviation log for QA to compare against the official record, with a written accuracy report at day 60. Starts within 45 days of signing; nothing replaces the official record.

the ask · $4,500 per pilot, then $900 per line per month at a founding price if the plant converts

a real yes · A real yes is a signed one-page pilot agreement, $2,250 received and a start date on the calendar; a QA head's enthusiasm, an unpaid 'bring us a demo' invitation, or 'come back after validation' are noes

3Small experiments

The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.

1. Fifteen QA head pilot offers
$1,300 · 18 days
Pull US contract packagers from DECRS and the Contract Pharma listings, attend one ISPE chapter meeting and the Contract Pharma conference, and book 15 calls with heads of quality. Founder pitches the $4,500 shadow pilot on one non-sterile packaging line in the first call, using a one-page scope, and asks directly whether a camera-drafted record may run alongside the official one.
keep going if · 5 of 15 calls end with a booked pilot scoping session and 8 of 15 name batch record review as a top-3 quality cost
2. Mock exception queue teardown
$700 · 10 days
Rent a GoPro, film a founder running a mock labelling and cartoning run against a standard master batch record template, then hand-write the batch record and deviation log from the footage alone. Walk 10 QA reviewers from the called plants through a side-by-side of a page-by-page review versus the exception queue.
keep going if · 6 of 10 QA reviewers say they would sign the exception queue in place of page-by-page review, and no more than 2 of 10 find a missed critical step
3. Close two paid pilots
$500 · 14 days
Send the one-page pilot agreement (60-day shadow mode, $4,500, half up front) to every scoped prospect, follow up twice by email, and drive to the two most engaged plants to sign in person. Invoice the $2,250 the day of signature.
keep going if · 2 of 15 offered plants sign the agreement and pay $2,250 within 35 days of the first call

4Collect a deposit up front

Tesla took $1,000 refundable reservations for the Model 3 and $100 for the Cybertruck before building either: the deposit is the measurement, not the revenue.

$2,250

per prospect, refundable

how · Paid design-partner pilot invoiced up front: the head of quality signs a one-page pilot agreement under their own approval authority and pays 50% of the $4,500 fee at signature, the standard instrument for an enterprise buyer who cannot buy an unvalidated product but can buy a bounded pilot set up: Stripe Invoicing ↗

what it reserves · One of three Q4 2026 pilot slots and a $900 per line per month founding price locked for year one after conversion

refund · Refunded in full if the pilot has not started within 60 days of signing or if we fail to deliver draft records for the first three lots

target · 2 paid pilots at $2,250 each from 15 offered plants within 35 days

before taking money · These are FDA-regulated GMP records: the pilot must run strictly in shadow mode beside the plant's official record, with no claim of 21 CFR Part 11 compliance or validation until a qualified consultant signs off.

Go: build it if

2 paid shadow pilots with money collected and start dates inside 45 days, plus 6 of 10 QA reviewers accepting the exception queue format

Kill: stop if

0 paid pilots after 15 offers, or 10 or more of 15 QA heads say a camera-drafted record cannot enter their quality system even in shadow mode, or every interested plant demands a bespoke validation package before any pilot starts

5 Scripts to run itoutreach message, landing copy, deposit terms · click to open

outreach message

Your QA team spends days per lot reviewing batch records for a packaging line where almost nothing goes wrong. I am building a clamp-on camera unit that drafts the batch record and deviation log, so review shrinks to an exception queue. Before building hardware, I am selling 60-day shadow pilots at $4,500: the camera output runs alongside your official record and never replaces it. Can I get 20 minutes this week to see if one of your non-sterile packaging lines qualifies?

landing page

Batch record review in hours, not days $4,500 buys a 60-day shadow pilot on one packaging line, with draft records delivered within 24 hours of each lot Book a 20-minute line qualification call

deposit terms

You pay $2,250 today, half the $4,500 pilot fee, which reserves one of three Q4 2026 shadow pilot slots and locks a $900 per line per month founding price for year one. The pilot runs 60 days alongside your official batch record and never replaces it. Full refund if the pilot has not started within 60 days of signing or if we fail to deliver drafts for your first three lots.

Would you run this test?

One tap. The yes-share feeds the Demand pillar of this idea's score; nobody sees who answered.

Budgets are out-of-pocket estimates for a team of one to three, US market. Size the deposit to the deal, and check the terms before taking money in a regulated line.

Scorecard

One score that balances how trendy the idea is, the demand for it and its potential for 100x, with competition measured relative to every other idea in the catalog. Recent startup trends first, government priorities second.

Idea Score, 0-100 · raw 47.2 x 1.61

Warm

competition: more crowded than 32% of ideas · headwind x0.84

+2.7

government priorities, secondary (15 matching grants)

Trend

Is the wave forming now? 2025-26 entrants vs 2023-24, rounds since 2025, the sector's live-batch direction, the 2026 trend analyst.

Entrants 2025-26 vs 2023-24 (similar companies)90
Rounds announced 2025+ in the sector25
Sector direction (live batch)100
2026 trend analyst25

Demand

Does anyone want it? YC's current RFS, companies already paid for something similar, the operator judge, founders' yes-rate in decks, readers who would run the test.

YC asks for it (current RFS: idea / sector)0
Someone already pays (similar companies, recent / all-time)50
Operator judge: real pain50

100x potential

Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.

Venture judge50
Market size axis67
Moat axis100
Neighbours still alive81
Technologist judge100

Score = 100 x cbrt(Trend x Demand x 100x) x (1 - 0.5 x crowding) + government bonus (max 5), calibrated so the 95th-percentile idea scores 90 (order never changes). A geometric mean: a weak pillar cannot be papered over. Percentiles are among the 272 ideas in the catalog; the terms matched were plug-in, camera, unit, writes, gmp, batch, records, humans.

The idea in full

What: Camscribe is a validated hardware unit that clamps onto a pharmaceutical or medical device production line, watches the operator steps and equipment displays, and produces the electronic batch record and deviation log that quality assurance staff currently type and double-check by hand. A plant orders units online per line, mounts them without an integrator, and pays per line per month; the unit ships with a pre-written validation package for 21 CFR Part 11 and Annex 11. Batch release is the cost being collapsed: days of documentation review per lot become a reviewed exception queue.
Why now: Federal money is moving into plant-floor reasoning rather than dashboards - NSF funded an I-Corps award for a neurosymbolic platform for root cause identification and corrective action in semiconductor manufacturing - while the current cohort attacks manufacturing with software only: Operon (yc S26) builds an agentic data layer for manufacturing and process industries and Orvixo (yc S26) does advanced planning for manufacturers, neither of which captures what happens at the bench.
Wedge: first customer and entry point: One contract manufacturing organisation running fill-finish lines, starting with a single non-sterile packaging line where the record is simple, the regulator's expectations are settled, and the QA head can approve a pilot without corporate procurement.
Path to 100x: Regulated manufacturing quality and compliance across pharma, device and food sits in the $10-100B band, and every validated unit adds labelled footage of deviations that makes the next customer's model accurate on day one and their validation package shorter. Shared deviation libraries across plants are the network: the vendor with the most installed lines has the only corpus a regulator has already accepted, and validation cost is what keeps a second vendor out.
Ceiling: If each plant insists on a bespoke validation, the product becomes a per-site engineering project and margins collapse to services.
Closest real companies, as the generator saw them: Operon and Orvixo sell manufacturing intelligence as software that depends on data plants do not have; Edviro operates energy infrastructure with AI but not regulated production records. Camscribe differs by generating the primary record itself in hardware, so the evidence trail is ours and validated.
Main risk: Validation and regulator acceptance take longer than the runway, and pharma quality leaders refuse to let a camera hold the primary record.

Five judges

Each judge scores every idea in the catalog with a named rubric; the venture judge decides whether a card is shown at all (4-5 is venture-grade).

Venture investor
3/5
A validated deviation corpus across plants could compound, but three years of R&D and bespoke per-site validation risks collapsing into services.
Bootstrapper
1/5
Validated hardware plus 21 CFR Part 11 acceptance means years of regulator review before a single line pays per month.
Operator
3/5
QA staff typing batch records is expensive and frequent, but a camera holding the primary record is a workflow and validation change quality leaders must approve.
Technologist
5/5
A validated camera unit reading operator steps and equipment displays produces the primary record itself, and each accepted validation package deepens a deviation corpus rivals lack.
Risk
2/5
A camera holding the primary GMP record needs regulator and QA acceptance that the card admits may outlast the runway, with per-site validation risk.
trends
2/5
A validated camera on GMP lines answers no 2025-2026 shift; the single NSF plant-floor award is secondary evidence worth at most one point.

Similar startups in the directory

Companies whose pitch matches most of the idea's terms (plug-in, camera, unit, writes, gmp, batch, records, humans): 5 all-time, 1 from the last two years. Same matching as Idea Check.

Klarifyyc X26 · 2026 · Vertical AI agentsalive
AI Agent for Therapists
Cozmo AIyc W22 · 2022 · Vertical AI agentsalive
End-to-end AI agents for property claims, from loss report to estimate
Veltyc W22 · 2022 · Agent infrastructurealive
SDK for adding an agent and human governance layer to your product
Agora.io500global · 2016 · AI infra and computeacquired
Agora Inc provides real-time communication solutions. The company offers real-time video calling, voice calling, live audio and video streaming, recording, and real-time messaging. It serves the gaming, retail, and education industries. The company operates in the People's Republic of China and the United States of America and the majority of its revenue is derived from the People's Republic of China.
東阪電子機器株式会社plugandplay · Robotics and physical worldalive
Tohan Electronic Equipment provides development, manufacturing and selling various types of motor drive unit of the numerical control device

Run this as an Idea Check →

The generator's reference companies

Real companies the model named as closest when it wrote the card, with their fate. A check mark is a company the radar could verify in its directory.

Public money in this direction

US federal grants, SBIR/STTR awards and open opportunities from the radar's public-money feed, matched to the idea's terms; the sector totals give the context.

grants and programs matching the idea

startup-relevant grants in Vertical AI agents

$135M

awarded in the sector, tracked

Bivalent CD47 Immunotoxin for Targeted Therapy of T-Cell Acute Lymphoblastic Leukemiaawardhigh relevance
NIH / NCI · STTR phase I · $400K · posted 2026-09-18
Phase 2: Development of an endocannabinoid microparticle formulation for the topical treatment of cutaneous manifestations of lupus erythematosusawardhigh relevance
NIH / NIAMS · SBIR phase II · $1M · posted 2026-09-25
Silk protein-based orthopedic implanted fixation devices for craniomaxillofacial applicationsawardhigh relevance
NIH / NIDCR · STTR phase I · $343K · posted 2026-09-23
A HERV-K (HML-2) Immunization Strategy for the Treatment and Prevention of Amyotrophic Lateral Sclerosisawardhigh relevance
NIH / NINDS · SBIR phase II · $497K · posted 2026-09-22
Dual Acting Bone Defenders for Inhibiting Cancer Mediated Bone Destructionawardhigh relevance
NIH / NCI · STTR phase II · $400K · posted 2026-09-18
Automated 2D Cell & Gene Therapy Manufacturing Platformawardhigh relevance
NIH / NIGMS · SBIR phase I · $221K · posted 2026-09-15
Miniaturization and Human-Centered Design of an AI-Enabled At-Home PKU Monitoring System for Childrenawardhigh relevance
NIH / NIGMS · SBIR phase I · $307K · posted 2026-09-05
Developing a broadly applicable targeted peptide-drug conjugate for the treatment of solid tumorsawardhigh relevance
NIH / NCI · SBIR phase II · $1M · posted 2026-09-04

All public money by sector →

Market signal

What the radar sees in Vertical AI agents: new companies by cohort year, the forming YC batch, and outcomes since the February snapshot.

Vertical AI agents · 159 → 261 → 364 → 411 → 353 new companies 2022 → 2026 · 93% aliveYC F26 live: 25 in this cluster, 20% of the batch (was 25% in S26)Since February, of 698 YC companies here: 19 acquired, 9 shut down, 219 rewrote their pitch

Vertical AI agents: companies, trend and grants →

Design attributes

The card is one cell of a designed set: every axis below was chosen before the text was written, and the text had to realize it.

Buyer: Enterprise
Business model: Hardware
Path to 100x: Collapses a cost 10x
Market size: $10-100B market
Capital intensity: Capital-medium (ops, field teams)
Speed to revenue: R&D first, revenue after 3 years
Technical depth: Real engineering
Go-to-market: Self-serve
Moat: Network effects
Geography: US first
Regulation: Heavily regulated
Vibe: Hot space