Theory Ventures

Explore the future of technology in our latest blog post on composable software platforms — learn how modern infrastructure and AI automation are setting the stage for software that adapts dynamically to business needs, enabling automation and efficient workflows. Interested in composable software? We'd love to connect at [email protected]! #ComposableSoftware #AI #Innovation

Launching our 2024 GTM Survey 📝

What does it take to go public? Has it changed over the last 15 years?

2012 was the year of the Seedpocalypse. Also called the Series A Crunch, a fear gripped Startupland : raising a Series A. Two years later, this indigestible excessive bolus of fundraising rounds hit the Series B market & Series Bs became the most challenging round to raise. Read More: https://lnkd.in/g2nRdupg

What stood out to me in yesterday’s Apple announcement wasn’t the headline, but the subtitle. “Setting a new standard in privacy.” For privacy to become one of the leading selling points of software, competitive dynamics & user preferences have evolved. The mantra repeated over the last 20 years on the internet has been privacy is dead. Users simply don’t care. People are willing to trade their privacy for free & targeted experiences. Since 2020, Apple has marketed their products as the private alternativepoking fun at how public we all are about the minutia of our lives. Apple is competing using the challenger sale : Apple is telling the market, to use AI, you should want private AI, you should want to control your own data because it’s valuable. Apple’s massive distribution will affect consumer & enterprise preferences. It’s already happening. Enterprises push for their data to be under their control in virtual private clouds, in open data formats like Iceberg, to run AI models in their own environments - the cloud-prem architecture. These models are their intellectual property & a key competitive advantage. We’ve all learned Google’s core search business - arguably the best business model on the internet - is the commercialization of a machine learning algorithm from user & enterprise data. Fool me once… The broader market has realized how valuable data can be - both users who provide it & businesses who leverage it. These changing preferences will create demand for new architectures : software products that run in customer environments where the data remains in control of the user & buyer. This architectural shift creates a massive new opportunity for software startups to challenge the multi-tenant, centralized data incumbents with their own privacy-focused challenger sale. Many will debate whether the architecture is truly private. Is Apple using secure enclaves & SGXs, OpenAI promising not to log queries from Apple users, & users opting-in to sending their data to OpenAI private enough? Will Google rearchitect Gemini to do the same thing to power Apple products? https://lnkd.in/g-ukCuGt

We are proud to lead Tobiko Data's $17.3 million Series A funding, driving advancements in data transformation frameworks for efficient and scalable data management. Excited to support Tobiko in revolutionizing how organizations handle data pipelines with SQLMesh and look forward to a future where data transformation is both simple and powerful with Tobiko Data.

We suffer from a cognitive bias : work performed by a human is likely more trustworthy because we understand the biases & the limitations. AIs are a Schrodinger’s cat stuffed in a black box. We don’t comprehend how the box works (yet), nor can we believe our eyes if the feline is dead or alive when we see it. New AI product on-boarding will need to mitigate this bias. One path may be starting with low-value tasks where the software-maker has tested exhaustively the potential inputs & outputs. Another tactic may be to provide a human-in-the-loop to check the AI’s work. Citations, references, & other forms of fact-checking will be a core part of the product experience. Independent testing might be another path. As with any new colleague, the first impressions & a series of small wins will determine the person’s trust. Severe errors in the future will erode confidence, that must be rebuilt - likely with the help of human support teams who will explain, develop tests for the future, & assure users. I recently asked a financial LLM to analyze NVIDIA’s annual report. A question about the company’s increase in dividend amount vaporized its credibility, raising the question : is it less work to do the analysis myself than to check the AI’s work? That will be the trust fall for AI. Will the software catch us if we trust it?

In our latest blog post, we're exploring the interaction between websites and AI agents — including our predictions for how web infrastructure will change to support AI agents. If you’re building a company around LLM-based agents, we’d love to hear from you at [email protected]!

For the past few days, I’ve been using the Mac ChatGPT app OpenAI demonstrated last Monday. It’s unquestionably the future of human-computer interaction. Conversing with a computer is much more natural than typing. Imagine speaking to a colleague with the entire internet at their disposal. But also a verbose colleague without much sense of social cues. Tapping a keyboard shortcut, the ChatGPT app loads & four little bars reminiscent of Google transcription software appear in the app. You can choose from 5 mellifluous voices who speak fluidly & naturally. The computer is patient with unstructured thoughts. I used it to outline this post. I switched between the strengths & challenges of the product randomly. The software categorized & organized, transforming the rambling mess of sentences to an outline. Verlyn Klinkenborg would still have some work to do, editing the outline. Just as most LLMs are verbose so is this app. For initial content review, that’s fine, but after the fifth or sixth iteration, it’s faster to interrupt the speaker by clicking. A hybrid voice & text mode would improve today’s either/or experience. Seeing the document evolving as the speaker chats would help with editing. Sometimes the conversation illusion breaks. The voice interrupts & says she couldn’t hear all of my sentences. Or the length of the input speech is too long & the app retries several times, processing smaller chunks of voice, leaving the user unclear on how much of the conversation has been captured. At least, this was my perception. Other times, the app isn’t patient with an um or a thoughtful pause. I imagine this a tricky problem : when is a pause an indication to speak or wait? But overall it’s clear where assistants like this will go : draft an email & send it. Delegate a task in Asana with a due date. Review a web page & summarize it in a blog post with some commentary, then check the grammar & publish it. All through voice. Years ago, I wrote The Dawn of the Voice-to-Text Era &  The Fastest User Interface on why I think voice is the future dominant user interface. We’re much closer to that vision than ever.

Join us on Thursday, May 30th, at 9:30 am PT for Office Hours with Tomasz Tunguz, who will be hosting Benn Stancil. Benn, co-founder of Mode, an analytics platform acquired by ThoughtSpot, writes extensively about the future of data and analytics. He explores deep questions about software, including "Do software companies actually have good margins?" and "Disband the analytics team." With a unique perspective on data and software, Benn will share his insights on significant topics, including BI’s third form. During this Office Hours session, Benn and Tomasz will discuss: -What is BI’s third form, and why it matters -The role of the analytics team in an AI-driven world where there are more shovels than gold -Margin structures for software companies: why they might be less attractive than they seem If you’re interested in attending, please register here: https://lnkd.in/gpgNfVQ3