Wordware: Revolutionizing Coding with Natural Language AI

Empowering Non-Technical Founders and Boosting Developer Productivity

This is The Startup Breakdown, the only newsletter looking into the new Y Combinator companies that you need to know. By joining this growing community of hundreds of founders, investors, and other startup aficionados (think i spelled that right?), you're getting a firsthand look at the companies that may be the next Airbnb, Coinbase, or Reddit.

Premium subscribers include investors and YC-backed founders looking to stay ahead of the latest startup news.

Wordware: The Future of Coding with Natural Language AI

Background:

If you’re good with your fingers → software

If you’re good with your hands → hardware

If you’re good with your tongue (that’s what she said) → Wordware

The proliferation of no-code and low-code tools has made it easier than ever for non-technical founders to build and gain traction on their ideas, and it’s easier than ever for someone without an engineering background (yours truly) to build.

However, whatever feasibility gain from the Bubbles and Glides of the world, the rise of AI might have a 10x improvement over the current generation of tools.

I outlined the process in last week’s newsletter on how I approach building with ChatGPT. Beyond general-purpose models, specialized tools are being built specifically for the next generation of coding.

  • side note, would be interested in seeing a correlation chart between chatgpt outages and github commits

Wordware is building one of these task-specific services, creating a natural language and collaborative IDE (think Microsoft Word for coding). It turns spoken words (I want a feature that does x) into functional code.

The appeal of natural language coding with Wordware is fairly obvious for non-engineers when it comes to why use a tool like this. Even for engineers, tools like this can speed up their production, allow them to better organize their code, and make it easier to onboard new team members.

Metrics:

Traction:

  • Customers including campfire, freeer.ai, Actaware

Market Size:

  • TAM: $4.5 billion

  • SAM: $900 million

  • SOM: $45 million

Competition:

  • OpenAI, Anthropic, Google

  • Rasa, Hugging Face

  • AgentGPT, GPTConsole

Team:

  • Filip Kozera, CEO: cofounded KRISTALIC (exited), which tried to augment the human memory, Deep Learning at Cambridge, Forbes 30 under 30 (uh oh…)

  • Robert Chandler, CTO: early engineer at a self-driving car startup, Five AI (acquired by Bosch), Information Engineering at Cambridge

Risks:

  • User experience: I consider myself to be a fairly technical person at this point, but even I was a bit overwhelmed with the tool’s function, where users must select loops, in-line code, and dozens of other features before typing out the request or objective. Even the explanations and demos on the company’s site left me with more questions than answers. The paradox of choice is a real thing, and this overwhelming interface could be an immediate objection for new users

  • No clear ICP: Who is Wordware built for? As with most founders, the team wants this product to be globally used. However, at least to start, it’s essential to build with someone in mind, and it’s really not clear who comes to mind for the team when they build new features and try to bring it to market

  • ChatGPT differentiation: As an avid ChatGPT dev, it’s going to be hard to convince me and the many like me to switch platforms unless there is some serious differentiation in the experience or product that Wordware can offer. Right now, I’m not seeing it

Awkward Season 4 GIF by The Office

Gif by theoffice on Giphy

What I like:

  • Experimentation: the platform provides examples of high-quality prompts and use cases, and it offers a variety of features that support experimentation to allow for rapid iteration and improvement

  • Quality assurance tools: Wordware provides an evaluation tool, can provide performance monitoring on specific parts of the prompt, and even supports version control to allow for experimentation 

  • API control: Wordware offers integrations with numerous APIs, is easily deployable for developers as a single API, and it even automatically switches the model being used based on each one’s status

Opportunities:

  • A/B testing: the experimentation features are already robust, but further leaning in and offering support for A/B testing of variable changes like prompt structure, word choice, and data provision even more useful for developer teams

  • Use cases and templates: as mentioned, the current site isn’t intuitive, and building out more use case examples and templates for quick implementation could move the needle for relevant decision makers looking for new dev tools for AI

  • Content engine: similarly, the opportunity to create educational content like how-to guides and free mini courses around agent building, and more specifically in NLE programming, could allow for better reach and discoverability from high intent users

 

 

Share this newsletter with one friend to get access to the Links of the Week section in every newsletter, including today’s! 👇️ 

  • Siri may finally become useful if Apple’s unveiled plans for the voice assistant hold true

    • The company announced a series of improvements at its annual WWDC (think Coachella for people who would genuinely choose their iPad over their newborn), and the general theme was one of integration

    • I (and many others) have long been annoyed with the lack of real interoperability of Siri with non-Apple apps such as Spotify as the OS has lacked the ability to effectively interact with third parties

    • Not only will this be solved (allegedly), but there is even a good chance that the voice assistant itself will become much more powerful as it is set to be powered by ChatGPT

    • I have been recently trying to use Siri again after not using it for literally 5 years, and though better than I remembered, it’s still noticeably limited

    • This strategy may be Apple’s moat in the AI space given their monopoly on the hardware space, and frankly, I’ve long believed that it is the only company capable of delivering on the whole “Jarvis” AI ambitions

    • And beyond Apple, this could have a massive impact on the broader AI hardware industry… 👇️ 

  • Humane is seeking a billion dollar acquisition after proving that slapping duct tape and a calculator together and putting it on a shoestring around your neck while calling it “AI” doesn’t automatically guarantee virality

    • The company, founded by former Apple-ers and funded pre-product for its promise of an always-on AI device that would dangle unobtrusively proved to be too much bark and not really any bite whatsoever, and after seeming to manufacture a narrative that such hardware was in hot demand, the final nail in the coffin was when tech-fluencer Marquess Brownlow infamously roasted the quality and user experience of the device, undoubtedly contributing to the devastating 10K units sold (or 10% of the forecasted goal of 100K…)

    • I am a true believer in the concept of proactive AI, and those seeking to build this technology have realized that owning the hardware is a requirement to bypass device risk

    • However, I still am skeptical of any new device pulling this off, even the Meta glasses, so long as Apple has the potential to turn the on switch and immediately crush any and every AI device company in existence

  • Photoshop’s new terms of service give Adobe permission to access projects even before they’re published for moderation purposes

    • There’s a chance this is in response to the rise of deepfakes and other altered images, particularly in the age of the election, but users aren’t thrilled, with some enterprise customers even reconsidering their licenses with the tech company over what they see as an infringement on what is considered sensitive information

    • An even more interesting point that could be brought up in relation to this: would this not make it possible for Adobe to be held accountable for harmful or illegally created content generated with Photoshop?

    • If they’re truly doing this to prevent this content in the first place, should they not then take accountability for this? Could something similar be extended to Meta and Instagram over content held in drafts?

    • Unlikely… this seems way too logical

  • The girls are fighting over in ecomm land as eBay ends its support for American Express cards

    • You can still use your friendly neighborhood Amex when you link to PayPal and pay this way, but the marketplace claims that the card issuer’s exorbitant fees aren’t seller-friendly…

    • This quickly caught the ire of the internet, as many pointed out that eBay’s own fees now frequently account for close to 20% of the sale price, so it’s not exactly a precedent for the company to come out pro-low fee

    • And in reality, I find it dubious at best that fees are the real reason behind the decision… American Express is notoriously consumer-friendly, saving internet purchasers from all sorts of scammy, sweatshop-made products and requiring rather extensive compliance work on the vendors’ parts in the process

  • Super interesting discussion on how image interpretation works with ChatGPT

Sponsored
Connecting Dots@dharmesh on startups, scaleups and strategy

Last word 👋 

How am I doing?

I love hearing from readers, and I’m always looking for feedback. How am I doing with The Startup Breakdown? Is there anything you’d like to see more or less of? Which aspects of the newsletter do you like most?

How disappointed would you be if you could no longer read this newsletter?

Login or Subscribe to participate in polls.

Cheers to another day,

gatsby

Reply

or to participate.