HyperWrite powers its generative AI service with Cohere

Why the language AI-first startup chose Cohere’s LLM

Today, the blank page is a thing of the past. Language AI innovators are building new and innovative ways to help writers of all kinds create the content they need to achieve their goals. One such pioneer is HyperWrite, a generative AI-powered writing assistant for consumers that makes the process easy and fast. The interface may be simple to use, but behind the scenes, a large language model (LLM) “puts pen to paper” and delivers unique and engaging text every time.

When choosing an LLM for their TypeAhead feature, the HyperWrite team considered several commercial and open-source solutions, as well as building their own. In the end, they chose Cohere’s generative LLM due to its high quality, low latency, and affordability. We sat down with Matt Shumer, Co-Founder and CEO of HyperWrite, to hear his perspective on partnering with Cohere for his fast-growing startup.

Meet HyperWrite

HyperWrite adds autocomplete capabilities to websites and apps, providing users with ideas and suggestions as they type. HyperWrite also enables users to enter prompts and generate complete copy for longer pieces such as emails, blog posts, business memos, and more. HyperWrite’s Chrome Extension is powered by Cohere.

HyperWrite helps anyone beat writer’s block and improve their writing process. Marketers use it to produce a drumbeat of marketing and ad copy. Students use it to generate ideas and deliver polished papers. Even English language learners use it to communicate more effectively at work or in their personal lives. HyperWrite’s fastest-growing segment is business professionals who are looking to be more efficient and productive when writing for any number of business needs, including emails, reports, and other business documents.

Under the Hood: Large Language Models

Previously, HyperWrite’s founders had been exploring the potential of large language models for some time. Shumer said, “Thankfully, these models really started to take off and it’s been a wild ride ever since (to say the least!). Now, they are so good that they enable things that we couldn’t have possibly imagined just two years ago.” He added, “And I think we’ll be saying the exact same thing two years from now. It’s all because of advances made by companies like Cohere.”

To deliver the best possible writing assistant, the startup wanted to build an interface that enabled the average person to interact with, and get value from, large language models (LLMs). Eventually, the HyperWrite team plans to move beyond writing to other applications that take advantage of LLMs.

Three Options: Hosted Solution, Self-Built, or Open Source

Shumer and team considered a variety of approaches to integrating LLMs into their platform, including building their own models from scratch or customizing an open-source solution. He said, “There are so many options out there, but at the end of the day, our first priority is our customers. By working with a managed solution like Cohere, we're able to focus our entire effort on our customers — understanding what they need, and building and iterating that for them.”

By removing the challenges of building their own solution, Cohere enables the startup to do more of what they do best and not worry about things outside of that. “Our entire business is language AI-first,” said Shumer. “It’s one of the most important functions we have to get right. It’s also quite complex and difficult to do, so managed solutions like Cohere help us implement language AI at a reasonable cost and with high quality.”

Partnering with Cohere

Quality is a theme that permeates HyperWrite’s experience with Cohere. The company tried working with a few managed language AI solutions before finally settling on Cohere for the long term. The deciding factors: accuracy, affordability, and great support. Shumer said, “Cohere’s models are amazing. The company excels at delivering cost-effective, low-latency models that are high-quality, which enables us to build some of our most difficult-to-implement features that can deliver excellent performance.”

During development, the Cohere team provided valuable technical support that helped the HyperWrite team achieve their goals. When they experienced problems or needed help figuring out a use case, Cohere expertise was right there for them. “That’s a superpower,” said Shumer. “Having Cohere’s team as an extension of ours enables us to do 10 times more, 10 times faster. They’re responsive over Slack, and always willing to try to push through boundaries for us.”

“Cohere’s models are amazing. The company excels at delivering cost-effective, low-latency models that are high-quality, which enables us to build some of our most difficult-to-implement features that can deliver excellent performance.”
~ Matt Shumer, HyperWrite, CEO and Co-founder

The HyperWrite Stack

HyperWrite is using Cohere’s Medium generation language model, which is a great fit for their product. It provides the perfect balance between speed and accuracy, enabling users to generate the output they expected within seconds, whether in the planning, drafting, editing, or iterating process with their piece.

When it comes to mitigating risk, Shumer is confident with his team’s approach. “Safety is a reasonable concern, but today’s LLMs have come a long way. Two years ago, you had no clue about what might come out of a model. Now, you have a good sense of how the model will behave.” HyperWrite also uses filtering technology built for social media, adding multiple layers of filters on top of their outputs. Text is not shown to the user until it’s checked, approved, and cleared.

So far, Shumer and team have seen very few problems, even with hundreds of thousands of monthly users. He said, “A lot of companies become too risk averse because they don’t realize that it’s possible to constrain things. Our approach has worked well for us and it’s also been easy to implement.”

Scaling Up at Warp Speed

HyperWrite is currently in hyper-growth mode, and their pace has been remarkable. “We’re actually outpacing our previous growth percentages and doubling our business every two months,” said Shumer. “And week after week, Cohere enables us to continue delivering at this ever-increasing scale, both from a platform and support perspective. We've worked with a number of vendors, and there's a clear difference in the quality of Cohere’s support, their ability to assist us and help us grow.”

Going forward, the team is excited to work on other applications as part of the paradigm shift towards AI-powered automation. They envision today’s personal assistant evolving to become more ambient, more personalized, and capable of helping people accomplish almost any task or everyday interaction. And LLMs will be core to this success.

For other business leaders looking to deploy language AI, Shumer has one piece of advice: get started today. “The companies that start now are going to be the ones that win in the long term because it’s all about learning how to make the best use of this powerful, evolving technology.”

To get started exploring the possibilities with language AI, create a free Cohere account.