Real-world data lab captures expert human demonstrations, builds RL environments from real workflows and evaluates agents with domain experts; first open benchmark on household manipulation due Q4 2026

SAN JOSE, Calif., Sept. 23, 2026 /PRNewswire/ — Realset AI (https://realset.ai), a real-world data lab that produces training data for LLMs, frontier models and embodied agents, today announced that Realset AI and Flatkey have raised $10 million in Series A funding. The funding will expand Realset’s capture network of real workplaces and studio environments, grow its pool of expert demonstrators and domain experts, and support open benchmarks that measure whether AI policies work outside the lab.

Why It Matters: The Internet Is Exhausted, the Physical World Is Not

Frontier labs and robotics companies have largely consumed the text available on the internet. The next gains come from data about people doing real tasks in real places: how a worker folds laundry, loads a dishwasher, packs an order or handles a customer return. That data does not exist on the web, and simulation does not reproduce it.

Realset’s answer is that every dataset starts with a real person doing a real task in a real place. The company uses expert demonstrators rather than crowd annotators, real environments rather than simulation, and delivers de-identified data to US-hosted cloud buckets, with quality measured at every step of a single pipeline.

Three Ways to Produce Ground Truth

Realset Body: embodied data for physical policy models. Egocentric and third-person capture of skilled workers performing manipulation tasks in homes, kitchens, warehouses and light assembly lines, plus bimanual teleoperation episodes with synchronized stereo video, IMU and action logs. Delivered with dense action-level annotation optimized for vision-language-action (VLA) training.Realset Field: LLM and agent training data from RL environments built on real workflows. Environments that mirror e-commerce operations, customer support, logistics dispatch and manufacturing SOPs. Domain experts generate trajectories, preference pairs and rubrics inside the environment, with verifiable rewards from real business outcomes and bilingual English/Chinese expert pools.Realset Judge: expert evaluation for agents in production. Evaluation design, failure diagnosis and continuous monitoring by people who do the job the agent is replacing, with targeted fix-data for the top failure modes and re-scoring as models and prompts change. The same services are offered to data integration and AI solution companies that deploy agents for their own clients.

The Realset Workspace

All three run on the Realset Workspace, where domain experts in six languages (English, Simplified Chinese, Japanese, Korean, Spanish and Arabic) answer real tasks, attach evidence and screen recordings, and pass an independent quality review before a record is approved. Every approved record exports as JSONL with provenance that customers can verify.

Open Benchmarks on Real Tasks

Realset publishes open benchmarks so the field can measure what matters: whether a policy works outside the lab. The Realset Household Manipulation Bench evaluates open-source VLA policies including π0, OpenVLA, GR00T and Octo on folding, loading, sorting and wiping tasks captured in real kitchens and laundry rooms, scored by success rate over three trials per task, with results expected in Q4 2026. A Light Assembly Bench, scored by the line workers who trained on the tasks, and a Commerce Ops Agent Bench for computer-use agents on real e-commerce seller operations are planned.

“The easy data is gone. What is left is the physical world, and you cannot scrape it,” said Hunter Guo, founder of Realset AI. “You have to put a camera on a skilled person doing real work, structure what they did, and check it with people who know the job. That is a capture and quality problem, not a labeling problem, and it is the problem we are building a company around.”

About Realset AI

Realset AI is a real-world data lab and training data provider for LLMs and embodied AI. It captures expert human demonstrations, builds RL environments from real workflows and evaluates AI agents with domain experts, for frontier labs, robotics companies, and data integration and AI solution providers. Realset is headquartered in San Jose, California. Learn more at https://realset.ai 

About Flatkey

Flatkey is an AI infrastructure platform that gives developers access to more than 100 official AI models and more than 1,000 AI tools through one key and one balance. Learn more at https://flatkey.ai 

Media Contact

Xingru Ren
Head of Marketing
+1 424 356 6176
xingru@flatkey.ai
https://realset.ai 

View original content to download multimedia:https://www.prnewswire.com/news-releases/realset-ai-and-flatkey-raise-10m-series-a-to-build-real-world-training-data-for-frontier-models-and-embodied-agents-302887825.html

SOURCE Realset AI

Longer video generation and richer reference inputs give small businesses new ways to develop marketing content, with savings determined by the work required to reach a usable result.

SINGAPORE, Sept. 23, 2026 /PRNewswire/ — MakeShot sees the strongest opportunity for small-team video marketing in reducing the work between a creative brief and usable footage. Capabilities such as longer video generation and richer reference inputs can help teams spend less effort assembling clips and correcting mismatched results, giving limited production budgets more room for creative testing.

For small marketing teams, production costs extend beyond filming or generating a clip. Preparing assets, matching shots, reviewing outputs, and revising details all consume resources. As AI video capabilities develop, the opportunity is to reduce these demands while giving teams more room to explore creative ideas.

Industry research illustrates the distinction between production gains and marketing outcomes. In Content Marketing Institute and MarketingProfs’ 2026 B2B research, 87% of marketers using AI for content creation reported improved productivity, while 39% reported improved content performance. These findings cover broader content creation, suggesting that faster production still needs to be paired with clear creative objectives and audience testing.

Longer Sequences, Fewer Assembly Tasks

Seedance 2.5, available through MakeShot’s AI Video Generator, supports videos of up to 30 seconds in a single generation. For a small team, that creates room to develop a connected sequence encompassing a product introduction, a setting, and a closing composition.

A skincare brand, for example, could explore a sequence moving from a moisturizer close-up into a morning bathroom scene before returning to the product. Generating that sequence together may reduce the work involved in assembling separate clips and reconciling differences in lighting, motion, or visual style.

The benefit depends on the result. A longer sequence that requires repeated regeneration can consume additional time and credits. Teams therefore need to consider how much of the output is usable and how easily it can be refined.

Clearer References, More Focused Creative Direction

On MakeShot, Seedance 2.5 supports up to 30 reference images, 10 video clips, and 10 audio clips, allowing teams to combine product, scene, motion, and sound references in one generation request.

For businesses with established product photography and brand assets, these inputs can communicate details that are difficult to express through text alone. Product images can guide appearance, scene references can establish atmosphere, and video or audio references can convey movement and sound.

MakeShot’s view is that the value lies in selecting relevant materials and assigning them clear roles. A focused set of references can provide a stronger creative brief than a larger collection containing conflicting directions. This gives small teams a practical way to apply their existing assets while retaining responsibility for product accuracy and final presentation.

Total Production Effort as the Measure of Efficiency

“For a small team, the most useful advance is one that removes work between the creative brief and the finished video,” said Wynn, Marketing Manager at MakeShot. “Longer generation and clearer references matter when they help teams reach an approved result with fewer revisions.”

MakeShot recommends evaluating total production effort, including unsuccessful generations, selection time, and final editing. Cost per usable video and time to an approved version provide more meaningful measures than the price or speed of a single generation.

This approach also helps teams decide where AI fits best. Concept exploration, atmospheric product footage, and creative variations may benefit from generation, while demonstrations requiring precise evidence of product performance may still call for filmed material. Allocating resources according to the task gives small businesses a clearer basis for managing production budgets.

About MakeShot

MakeShot is an AI video and image generation platform that gives creators and businesses access to multiple generative models through a browser-based interface. Its AI video generator supports text- and reference-driven creation for social media content, marketing projects, and visual storytelling.

Media Contact:
Wynn
Marketing Manager
Email: support@makeshot.ai
Website: https://makeshot.ai/

View original content:https://www.prnewswire.com/news-releases/from-product-photos-to-campaign-content-makeshot-on-ai-video-efficiency-for-small-teams-302887829.html

SOURCE MakeShot

The platform replaces a growing stack of provider accounts, credits and keys with one key, one balance and one invoice, at 60–90% of official list prices

SAN JOSE, Calif., Sept. 23, 2026 /PRNewswire/ — Flatkey (https://flatkey.ai), the AI infrastructure platform that brings models, tools and data together behind one API key, today announced that Flatkey and Realset AI have raised $10 million in Series A funding. The company also said that more than 10,000 developers have adopted the platform in the two months since its July 2026 launch, using a single API key and a single balance to access more than 100 official AI models and more than 1,000 AI tools. The Series A funding will go toward adding more official models and tools to the platform and scaling the infrastructure that routes developer traffic to them.

Flatkey is built as production infrastructure for AI developers, not a convenience layer. Every call is routed to the provider’s official endpoint, and because Flatkey buys upstream capacity in volume, most models are priced at around 80% of the providers’ official list prices, with some at 60% or lower as part of ongoing promotions. Developers pay less than they would going direct, and they do it through one key and one balance across models and tools.

Why It Matters: Models, Tools and Data Are Becoming One Layer

AI is shifting from answering questions to completing work. The agents doing that work depend on three things: the models that reason, the tools that act, and the data they act on. Each of those is fragmenting into more providers every quarter, and every provider adds an account, a top-up, an API key, a rate limit and an invoice.

Flatkey’s bet is that these three converge into a single layer developers reach through one key. A production application today rarely depends on one model: it combines several models across text, image, audio and video and pairs them with tools such as search, browsing and data enrichment. A team using models from OpenAI, Anthropic, Google and DeepSeek, a video model such as Seedance, and a search and a browser tool would ordinarily manage seven or more provider relationships. With Flatkey that becomes one key, one balance and one invoice, and a new model or tool is a parameter change rather than a new vendor.

What Developers Get

More than 100 official models from OpenAI, Anthropic, Google, DeepSeek, Kimi, GLM and others, plus image and video models such as Seedance. Every call is routed to the provider’s official endpoint. Flatkey does not self-host modified or quantized versions and label them as the original model.More than 1,000 tools on the same balance: search, browsers, data enrichment, media generation and actions, with no separate billing setup per vendor.Simple pricing. Subscription plans start at $10 per month, and pay-as-you-go credits cover both models and tools on one balance.A one-line migration. Flatkey is a drop-in replacement for any OpenAI-compatible client: developers change the base URL and their existing code works.New models on release day. Flatkey adds new models through official channels as soon as they are released, so teams do not open a new provider account every time something new ships.

“AI development is becoming less about choosing one model and more about combining models, data and tools across text, image, audio and video,” said Hunter Guo, founder of Flatkey. “If developers can reach all of that through one key, the platform stops being a convenience layer and starts to look like infrastructure. That is the company we are building.”

Availability

Flatkey is available today at https://flatkey.ai. New users start with $1 in free credit and can continue with pay-as-you-go credits or a monthly plan.

About Flatkey

Flatkey is an AI infrastructure platform that gives developers access to more than 100 official AI models and more than 1,000 AI tools through one key and one balance. Headquartered in San Jose, California, Flatkey launched in July 2026. Learn more at https://flatkey.ai 

Media Contact

Xingru Ren
Head of Marketing, Flatkey
+1 424 356 6176
xingru@flatkey.ai
https://flatkey.ai 

View original content to download multimedia:https://www.prnewswire.com/news-releases/flatkey-raises-10m-series-a-after-surpassing-10-000-developers-in-two-months-302887820.html

SOURCE Flatkey