Sublingual

Overview

Sublingual was a two-person Y Combinator Winter 2025 company founded by Matthew Tang and Dylan Bowman in San Francisco. Despite being listed in the YC directory as a "daily productivity tracker" — almost certainly a placeholder or relic of an earlier concept — the company actually built an open-source, local-first LLM observability and evaluation platform. Its core pitch was radical simplicity: a single pip install subl command that required zero code changes and stored all logs locally, targeting developers who were skipping rigorous LLM evaluation in favor of intuition and "vibe tests."^[1]

Sublingual failed because it entered a space being simultaneously crowded by well-funded competitors and commoditized from above by the very LLM providers whose outputs it was designed to monitor. Its free-forever, open-source model maximized developer accessibility but left no monetization path that a two-person team with $500,000 in runway could survive long enough to reach.

The company is now listed as "Inactive" on the YC directory, with both founders designated "Former Founders."^[2] No acquisition, acqui-hire, or public post-mortem has been recorded. The entire arc — from YC acceptance to inactivity — appears to have played out within a single calendar year.

Founding Story

Matthew Tang and Dylan Bowman brought complementary, directly relevant backgrounds to Sublingual. Tang had worked at TikTok on machine learning for the recommendation algorithm and ads engine, then moved to Nextdoor where he conducted LLM research for recommendation systems.^[3] Bowman came from academic and government research: he had conducted LLM research at the UIUC Kang Lab and at the U.S. Department of Defense.^[4] In November 2024, just months before founding Sublingual, Bowman co-authored a paper at UIUC demonstrating that voice-enabled AI agents could autonomously execute financial scams using ChatGPT-4o's real-time voice API — work that put him in direct contact with the failure modes of deployed LLM systems.^[5]

The founding insight was practical rather than theoretical. Both founders had spent years building and evaluating LLM applications and had watched the same pattern repeat: developers, under pressure to ship, would skip structured evaluation entirely. As the founders described it on their YC profile: "We've spent years building and researching LLM applications, and we've seen firsthand how developers handle evaluation: sifting through logs, relying on intuition, and struggling with the friction of integrating existing observability tools when they just want to focus on building."^[6]

They validated this observation through founder conversations: "Through conversations with numerous founders, we've learned that they're often too busy building to establish robust evaluation systems. So, they end up relying on a couple vibe tests before crossing their fingers and pushing to prod."^[7]

The company's name encoded its philosophy. "Sublingual" — the medical term for under-the-tongue drug delivery, chosen for its implication of fast, frictionless absorption — was a deliberate signal that the product would dissolve into a developer's workflow rather than demanding a new one. The YC launch post title, "LLM evals for lazy devs," reinforced the same message: the product was designed for developers who would not adopt anything that required effort.^[8]

One notable gap in the founding story is the discrepancy between the YC directory description — "Daily productivity tracker" — and the actual product built. Whether this reflects an early pivot away from a consumer productivity concept, a deliberate misdirection during the application process, or simply an uncorrected placeholder is unknown. No founder statement addresses it directly.

Timeline

• November 2024 — Dylan Bowman co-authors UIUC research paper on AI voice scams using ChatGPT-4o, establishing his LLM research credentials immediately before founding.^[5]
• January 2025 — Sublingual accepted into Y Combinator Winter 2025 batch.^[1]
• March 3, 2025 — Sublingual featured in a developer tools roundup article, the only press coverage found, describing the product as an open-source LLM observability and evaluations platform.^[9]
• March 12, 2025 — Sublingual closes $500,000 pre-seed round from Y Combinator — the sole funding event on record.^[10]
• Early–Mid 2025 — Y Combinator promotes Sublingual AI on LinkedIn as "plug-and-play LLM evaluations" during the active batch promotion period.^[11]
• 2025 (exact date unknown) — Sublingual listed as "Inactive" on YC directory; founders listed as "Former Founders."^[2]
• 2025 (post-Sublingual) — Matthew Tang's LinkedIn shows activity on new projects, including an open-sourced Cluely-like product and a project called "zbench," indicating the founders moved on to new work.^[12]