Refresh C Verified
RL environment company · Coding
- Domains
- Coding Computer Use Private Codebases
- On X
- @refresh_dev ↗
- Funding
- Y Combinator X25 · rounds and investors
- Tier
- C why
- Snapshot
What Refresh does
Refresh builds RL environments and evals with verifiable rewards for coding, computer use (cloned software suites like EHRs and enterprise apps) and 3D animation in Blender, for frontier labs and enterprises.
- Sells
- RL environments and Data
- Leads with
- Several domains (a generalist)
- Based in
- San Francisco, USA
- Team
- 1-10
- Website
- refresh.dev ↗
What Refresh sells
Read from Refresh's own website, checked October 12, 2026: the line that puts it in each category.
RL environments
“Refresh builds simulated worlds for coding and computer use, with verifiable rewards that move models forward.”
refresh.dev ↗“Refresh builds these environments as both evals, to objectively benchmark how well frontier models perform real computer work”
refresh.dev ↗
Data
“Email contact@refresh.dev to request environments or discuss a bespoke dataset”
refresh.dev ↗
Domains Refresh covers
“Software engineering work that targets GDPval capabilities, the economically valuable tasks professionals handle across dozens of occupations.”
refresh.dev ↗“High-fidelity clones of full software suites, wiring together multiple connected applications, from EHRs to enterprise apps and beyond.”
refresh.dev ↗“from EHRs to enterprise apps and beyond”
refresh.dev ↗“from EHRs to enterprise apps and beyond”
refresh.dev ↗“Agents working inside Blender, from building scenes and materials to rigging characters, keyframing animation, and running cloth and physics simulations.”
refresh.dev ↗
Refresh competitors and alternatives
The companies that sell the most of the same things in the same domains, and what they share with Refresh.
- AfterQuery Also sells RL environments, datasets and evals · also in coding, computer use and enterprise work
- Collinear Also sells RL environments, evals and datasets · also in coding, computer use and enterprise work
- ReasonCore Also sells datasets, RL environments and evals · also in health and bio, coding and enterprise work
- Surge AI Also sells RL environments and evals · also in coding, computer use and health and bio
- Theta Also sells RL environments and evals · also in coding, health and bio and enterprise work
Tier C
Early: a small team.
How much stands behind a tier: usage, who backs it, a confirmed line from its own site or paper, and a clear read on X. High with three of the four, medium with two, low otherwise.
- team 1-10
- investors: Y Combinator
- went through Y Combinator Spring 2025
- in curated RL directories (5 pts)
On X · unclear, low buzz, from 12 posts about Refresh in the last month
A small circle of replies praises Refresh's newly released Blender/animation benchmark results, and one VC flags it as an RL gym company; little broader attention.
How we got to tier C
- Adoption: team 1-10; investors: Y Combinator; went through Y Combinator Spring 2025. Tier C.
- On X: unclear, low buzz. “A small circle of replies praises Refresh's newly released Blender/animation benchmark results, and one VC flags it as an RL gym company; little broader attention.” No move.
- In curated RL directories: 5 points, a floor of C.
- Confirmed: a line from its own site or paper places it.
A clear view on X moves a tier one step up or down; most entries don't move. How tiers work.
Refresh funding
- Accelerator
- Y Combinator X25
Backed byY Combinator Spring 2025
Accelerator
- Y Combinator Spring 2025
- Y Combinator's company directory lists Refresh with the website refresh.dev, in its Spring 2025 batch
- Checkable on YC's page: ycombinator.com/companies/refresh
- No other page read names Y Combinator yet
Every figure is quoted from its own site, its accelerator or a funding announcement; a round's amount and lead appear only when its quote states them. Startup databases are never used as evidence. Checked 2026-10-11. How funding is checked.
Training methods
Methods Refresh uses, offers or builds for, with public evidence: its own website says so, or two independent sites do.
- RLVR Reinforcement Learning with Verifiable Rewards On its own website
“An RLVR dataset curated from Gauntlet, delivered in Harbor format for reinforcement learning.”
refresh.dev“\\ \\ Datasets\\ \\ Feb 17, 2026 Gauntlet 4K RLVR \\ \\ An RLVR dataset curated from Gauntlet, delivered in Harbor format for reinforcement learning.\\ \\ Read more”
refresh.dev - GRPO Group Relative Policy Optimization On its own website
“We trained rnj-1-instruct 8B using the terminus-2 harness with our Gauntlet 4K RLVR dataset. After GRPO showed instability during training, we switched to the DAPO algorithm which provided consistent convergence.”
refresh.dev - SFT Supervised Fine-Tuning On its own website
“…k, and as training gyms with verifiable rewards that models learn from. A verifiable reward is an objective signal of whether the agent completed a task correctly, and it powers the two main ways to post-train an agent: supervised fine-tuning (SFT) on high-quality example trajectories, and reinforcement learning, where the agent improves by acting in the environment and being scored on the outcome. That signal can come from a deterministic verifier, such as…”
refresh.dev
Where Refresh sits
Its domain on the map, live. Hover a logo for its name; click to open it. Open it on the full map
Latest from Refresh on X
@refresh_dev ↗Next: browse all environments. Entry from the RL Research daily scrape of public sources, 2026-11-05 snapshot.