YC Spring 2026: First Durability Assessment of 203 Startups
Of 203 YC Spring 2026 startups assessed July 2026, 124 landed losing its protection and only one earned a positive standing. Here is where the pressure comes from.
WillItSurvive is a research site that assesses company durability at scale. Its July 2026 dataset covers 10,525 companies, each researched individually and checked against public sources. The method was tested against 116 companies that have already shut down, using only information that was public 18 months before each one closed: 90% of those companies were flagged as at risk, and none were cleared as safe. We never declare an outcome for any living company; everything forward-looking here is a dated watch-item you can check yourself.
This cut looks at one cohort inside that assessment: the Y Combinator Spring 2026 batch, 203 companies assessed in July 2026, most of them only months old.
The headline numbers
The distribution is stark. Of the 203 YC Spring 2026 companies assessed in July 2026, 124 (62%) landed in the "Losing its protection" standing, 44 were rated "Little protection left," 34 "Protection under test," and exactly one, Harbor, earned "Defensible for now." Across the batch, competitive pressure ran consistently high: more than a quarter of the 203 companies were assessed as facing a direct, well-funded rival already active in their niche. Twelve companies in the batch were assessed as facing the most direct competitive exposure in July 2026: CharacterQuilt, Gravy, AndCo, Superlog, CentralComs, Ara, Allowance, Result, Keyframe Labs, Saffron, jo, and Replicas.
That is not a harsh grading curve applied to young companies. It reflects a specific pattern: a lot of founders built in the same three months, on the same foundation models, watching the same demos, and walked into rooms that were already full.
Where the pressure comes from
Five recurring squeeze points show up across the batch.
Coding-agent wrappers. Ara, Replicas, Runtime, and Hyper all face direct pressure from Anthropic's Claude Code, OpenAI's Codex, GitHub Copilot, and Cursor, per the July 2026 assessment. The model makers and established dev tools are shipping the exact orchestration, context, and governance layers these startups were building as a wrapper. Ara, a two-person team building a coding IDE on top of Claude and Codex, was among the twelve companies assessed as most exposed in the batch.
Counter-drone defense. Three companies in the batch, Surtr Defense Systems, Surtr Defense, and 9 Mothers, are building counter-drone systems. The dominant pressure on all three comes from Anduril's Lattice platform and Picogrid, citing Anduril's reported $20B Army deal and Picogrid's Northrop Grumman partnership (as reported through July 2026; verify current contract status independently). The concern is not the startups' technology but budget lock-up before any of the three lands a first paying customer.
Sales, marketing, and website agents. CharacterQuilt, GojiberryAI, and Ploy sell agent tools that HubSpot, Salesforce, and Webflow are building directly into products customers already pay for. When there is no separate purchase to make, the wedge closes.
Commerce agents. Kinect and Light Anchor face better-funded rivals such as Rep AI plus Shopify's own agentic commerce tools, which reach the same shoppers with native distribution.
Consumer finance and research. Gravy, Standard Signal, and Cohesion were assessed in July 2026 as facing pressure from OpenAI's ChatGPT Finances, Perplexity's Plaid integration, and established quant shops like Numerai and Voleon.
The narrow openings that held up better
A small group stood out favorably, and the common thread is not superior technology. It is picking a workflow that no giant platform already owns.
Harbor is the standout. In the July 2026 assessment, Harbor was the only company in the 203-company YC Spring 2026 batch to earn a positive standing, "Defensible for now," on the strength of 7 clinical trials run and a $1.93M contract booked while pulling sponsors away from IQVIA, ICON, and Parexel. It also showed far less competitive exposure than the rest of the batch.
Four others faced comparatively little competitive pressure, for stage-appropriate reasons. Trellis (short-term rental operations) has no incumbent breathing down its neck yet, though the assessment notes this partly reflects that it is only two months old. Dayjob sits in waste-trucking dispatch, where Samsara is an adjacent threat rather than an active competitor today. Nine Fives holds a defensible niche in RF test equipment for aerospace and drones, with no evidence yet of a well-funded rival closing in. Final Dose also showed low exposure, with a narrow cancer-drug approach that lacks an obvious well-funded rival so far.
The sector-level read from the assessment: the opening is real when it is narrow and unglamorous. The wide-open finance, avatar, and app-builder ideas mostly are not.
Cautionary cases inside the batch
A few named examples illustrate why 62% of the batch landed losing its protection in July 2026. CentralComs is a two-person team syncing to property management software via CSV while AppFolio and Buildium build the same maintenance and leasing agents natively. Superlog sells an auto-fix-PR workflow that Datadog and Sentry already ship inside tools engineering teams already pay for, with the better-funded Resolve AI chasing the same buyers. Result entered the AI app-builder category pre-launch, with no paying customers, against Lovable, Bolt.new, and Wix's Base44. Keyframe Labs sells a real-time avatar API against Anam, Tavus, HeyGen, and Synthesia while Google bundles avatar generation into Workspace. None of this settles any outcome; it explains the competitive pressure described here.
Questions this data answers
Which YC Spring 2026 startups scored best on durability? Harbor was the only company in the 203-member batch to earn a positive standing in the July 2026 assessment, backed by 7 clinical trials and a $1.93M contract. Trellis, Dayjob, Nine Fives, and Final Dose also faced comparatively little competitive pressure, mostly because their niches lack an active large rival today.
Why did 124 of 203 companies land losing its protection? Most are building thin layers on top of tools that OpenAI, Anthropic, Salesforce, HubSpot, or Shopify are now shipping natively, as found in the July 2026 assessment. When the underlying platform ships the same feature for free, there is no separate purchase left to make.
How crowded is the coding-agent space in this batch? At least four companies, Ara, Replicas, Runtime, and Hyper, were assessed as facing direct pressure from Claude Code, Codex, GitHub Copilot, and Cursor in July 2026. The model makers themselves are shipping the orchestration and context features these startups sell as standalone products.
Which counter-drone startups came out of this batch? Surtr Defense Systems, Surtr Defense, and 9 Mothers, and all three were assessed in July 2026 as squeezed by Anduril's Lattice and Picogrid, whose reported Army and Northrop deals are locking up budgets. None of the three had a documented paying customer at assessment time.
What is the healthiest niche in the batch? Clinical trial operations, based on Harbor, the only company holding real revenue and moving sponsors away from legacy CROs as of July 2026. Waste-trucking dispatch (Dayjob) and short-term rental operations (Trellis) also showed less competitive pressure than most of the batch.
What to watch
These are checkable items, not conclusions.
- By March 2027: whether GitHub Copilot, Cursor, or Claude Code ships a remote cloud-agent orchestration feature mirroring what Replicas or Runtime sell standalone.
- By June 2027: whether Anduril or Picogrid announces another counter-drone contract overlapping the market Surtr Defense Systems, Surtr Defense, or 9 Mothers target.
- By December 2027: whether Harbor reports more than 15 total clinical trials, roughly doubling its July 2026 count of 7, and whether Harbor or Dayjob discloses additional paying contracts, which would show the batch's narrow niches can convert to revenue.
- Ongoing: any announcement by HubSpot, Salesforce, Shopify, Snowflake, Databricks, or Webflow of a native agent feature matching one of these startups' core function, which would mean that startup's wedge has closed.
Assessments reflect evidence available July 2026. Companies change; treat watch-items as things to verify, not conclusions.
Figures are as of each company's assessment date unless dated otherwise.