Mintlify is building a next-gen platform for writing software docs

by with No Comment Tech

Software developer coding with code coming out of laptop in futuristic looking way.

Image Credits: Kriangsak Koopattanakij / Getty Images

Software documentation — the resources that explain how software works and how to use it — has evolved dramatically over the past few decades. Once mostly in the form of PDFs and static plaintext, docs today are more interactive and user-friendly than they used to be.

But crafting them is still a time sink. Han Wang and Hahnbee Lee, both devs and entrepreneurs, say that they’ve personally struggled with this.

“In the 2010s, companies like Stripe, HashiCorp, Twilio and a bunch others raised the bar on developer content,” Wang said. “They proved that a truly great developer experience for their content is not just a commodity, but a competitive advantage. Since then, every company has been trying to catch up, but it’s actually quite difficult.”

Inspired to try simplifying the workflows to publish docs (mainly to make their own lives easier), Wang and Lee created Mintlify, a collection of documentation-authoring tools, including tools that can auto-generate docs from codebases.

“In the 2020s, the bar for documentation is rising once again,” Wang said. “This time, it’s not just with the UI, but the way developers and editors are fundamentally interacting with the content because of AI.”

An AI-powered vision

Wang and Lee met in college. The two attended Cornell; Lee was a computer science undergrad, and Wang was studying for a bachelor’s in information science.

As a student, Wang founded two companies: FoodFul, a system for remotely monitoring livestock, and People, a platform to build customer communities. After People (which Lee helped found) was acquired by user engagement firm Bettermode, Wang stayed on for several months but eventually left to become a partner at Bain Capital Ventures.

Wang left Bain in 2021, which happened to be right around the time he and Lee had the idea for Mintlify. They raised seed capital from Bain (leveraging Wang’s connections) and others, including Sourcegraph co-founder Quinn Slack, to grow the platform into a business.

At a high level, Mintlify assists devs with writing guides, API references, SDK docs and chatbots (powered by OpenAI’s API) to explain the ins and outs of their software and services. It provides built-in components and templates for basic doc formatting, and structures docs so that they can be embedded in a codebase.

Mintlify
Mintlify provides tools to write and maintain documentation for software, including tools that auto-update docs.
Image Credits: Mintlify

To help maintain docs, Mintlify also routinely scans for “stale” documentation and detects how users are engaging with the content to suggest ways to improve readability.

But there is some criticism of Mintlify’s automation features.

An early user, DevClass’ Tim Anderson, claims Mintlify adds comments to codebases that are of “little value,” and in one instance repeated the same factually wrong sentence in a doc four times over. Others have pointed out that Mintlify can be confused by disorganized and unoptimized, or otherwise poorly written, code.

Wang emphasizes the potential of the platform’s AI over its limitations, while implying that humans can’t be removed from the documentation-writing loop entirely.

“As we saw it, the role of content was changing with AI. Documentation will evolve automatically in real time from support messages, the codebase and product feedback,” Wang said. “AI assistance will help companies write technical content automatically based on product changes and user feedback.”

An expanding business

Mintlify isn’t the only startup trying to revamp how devs make and publish technical guides.

There’s Guidde, whose AI automatically generates software documentation videos. More in line with what Mintlify’s doing, Documatic automatically produces changelogs and explanations from code in addition to documentation.

I mentioned rivals to Wang, and he responded by highlighting Mintlify’s rather impressive customer list, which includes Anthropic, Cursor, Perplexity, Zapier, Polymarket, Fidelity and around 3,000 other brands. (Wang estimates that Mintlify’s tools reach more than 1.5 million developers a month.)

He also hinted at differentiating capabilities coming to the Mintlify platform in the near future.

“Every documentation now needs to have an AI chat to answer questions directly. But it will go much deeper into that,” Wang said. “Content creation will also change … The content will be used to power support, chatbots and generative AI models themselves. Content will also be personalized for every reader.”

To make that vision a reality, Mintlify recently closed an $18.5 million Series A funding round led by Andreessen Horowitz with participation from Bain and Y Combinator. (Andreessen Horowitz general partner Jennifer Li is joining Mintlify’s board as a part of the deal.) This brings Mintlify’s total raised to $21.7 million; Wang says that the new cash will be put toward expanding Mintlify’s 11-person team and product development.

“We have always focused on operating lean and efficiently,” Wang said. “We did not need to fundraise, but strategically decided to in order to fuel further growth.”

Wang declined to answer questions about Mintlify’s revenue and profitability.

Hello Wonder is building an AI-powered browser for kids

by with No Comment Tech

Hello Wonder app displayed on smartphone and computer screens

Image Credits: Hello Wonder

Across the world, regulators have ramped up their efforts to try and increase the safety of kids on the internet. Major social networks are facing scrutiny, and as a countermeasure, trying to roll out tools to protect kids. The core issue in the focus is the content that shows up on children’s screens and how to make it safe.

While a lot of these efforts are geared toward teens, toddlers also use devices to consume content. So, a trio of founders who have worked at companies like Google and Amazon are trying to create an AI-powered browser/companion to create a safe environment for kids to learn and explore through Hello Wonder.

The company currently has an iPad app — which parents have full control over — that lets kids ask questions to an AI chatbot and get answers, videos and interactive experiences that are safe for them. The startup believes that current content tools like YouTube Kids are focused on more engagement and don’t give parents enough insights about what their kids are consuming. That’s the problem the company has set out to solve.

Hello Wonder has raised $2.1 million from investors such as Designer Fund, a16z Scout Fund, Ground Up Ventures and Chasing Rainbows. Investors also include individuals like kids’ content studio PocketWatch’s CEO Chris Williams, Things, Inc. founder Jason Toff and electronics-focused fund MESH’s CEO Tony Fai.

Hello Wonder was founded by Seth Raphael, who led AI prototyping teams at Google and helped built the first version of Google Photos; Brian Backus, who worked as a games producer at Amazon, Disney, DreamWorks and NBCUniversal; and Daniel Shiplacoff, a product designer who worked on Google’s Material Design guidelines.

Raphael built the app out of necessity while raising five children under 12 during the COVID-19 pandemic. He told TechCrunch that while he had seen the potential for AI to help children while studying at university, the technology wasn’t ripe.

“The fundamental problem is that you and I use the internet wonderfully every single day and get tremendous value out of it. But we can’t let our kids do that because there is real harm. Plus, young kids don’t have the ability or tools to find out the content that’s helpful to them,” he said.

Image Credits: Hello Wonder

Raphael said that he began by trying to find the best content for his children. But that was constraining when kids wanted to explore a certain topic more. Then, he took inspiration from the Montessori method of learning, which involves hands-on learning and activities based on children’s interests. That led the company to build an AI-powered environment to bring content from different corners of the internet in a safe manner.

The company lets parents control what kind of content — videos, games and material from websites — their children are consuming. They can get texts about all kinds of videos or choose to get a daily or weekly summary of their consumption. Parents and guardians can tell the AI through the parent interface in natural language about the content they want and don’t want their children to consume.

For instance, if a family wants to help their kid learn violin, they can tell Hello Wonder that, and the tool will find and insert content about learning violin from time to time.

Hello Wonder, which targets kids from ages five to 10, also lets them interact with trusted family contacts through messages and video calls within the app.

Jordan Odinsky, a partner at Ground Up Ventures, said that Hello Wonder solves the problem of kids having to see unsafe content by involving an AI and having it scan content for safety before serving it to kids.

“Safety systems on today’s apps services for children don’t go far enough. As a browser, Hello Wonder doesn’t lock kids into any one format. They’re free to explore with the AI watching over them. They can consume any type of content as long as it fits within the parent’s values giving them a true internet experience,” he told TechCrunch over a call.

Odinsky added that the app could also be adopted as the child matures and show content to reflect that growth. He said that the app doesn’t have a problem presenting children with a blank search box and leaving them clues about what they want to ask.

“Wonder is built differently. When kids log on, they are prompted each time with ideas to search for. From there, it sparks new ideas to explore that you simply input by speaking. Many of the things browsers deal with, from exploration to discovery to figuring out the best prompt to achieve a desired result, are removed from the Wonder experience,” he noted.

The company is not charging a fee for the app at this time but will introduce a subscription layer in the future. It is also testing to expand the app to Android tablets and Chromebooks.

TechCrunch Space: Building (and testing) for the future

by with No Comment Tech

Image Credits: TechCrunch

Hello and welcome back to TechCrunch Space. It’s becoming a habit to open each TechCrunch Space newsletter with a bit of an update on Boeing’s Starliner mission, so bear with me. Per NASA officials, the spacecraft will now return its two-person crew to Earth no earlier than June 26, instead of the originally planned date of June 14. In other words, they’ll be spending at least 20 days on board the International Space Station instead of just eight. 

Read my story from last week on the causes of the additional delay and what it might mean for Boeing’s Starliner program.

Want to reach out with a tip? Email Aria at [email protected] or send me a message on Signal at 512-937-3988. You also can send a note to the whole TechCrunch crew at [email protected]. For more secure communications, click here to contact us, which includes SecureDrop instructions and links to encrypted messaging apps.

Story of the week

The space industry is all abuzz about how SpaceX’s Starship, Blue Origin’s New Glenn and other heavy-lift rockets will change just about everything. One likely consequence is that spacecraft will get bigger — much bigger — as engineers work outside the constraints of low mass requirements. 

There’s one problem: The current testing regime for spacecraft is focused on payloads four meters or less across. Gravitics and NASA are looking to change that, with a new agreement aimed at addressing this dearth of testing and qualification methods for larger spacecraft. 

Image Credits: Gravitics (opens in a new window)

Scoop of the week

Like many highly valued startups, SpaceX sometimes allows its employees to cash out some of their shares by selling to company-authorized outside investors.

TechCrunch has gotten a peek at an internal SpaceX document about such a tender offer from May 2022. Musk posted on X last month that SpaceX holds such sales for employees about every six months.

These documents offer interesting insights into the investors who are authorized to buy these secondary shares, and the good deals they get. Click the link above to take a look.

South African businessman Elon Musk arrives at the Tenth Breakthrough Prize Ceremony at the Academy Museum of Motion Pictures in Los Angeles, California, on April 13, 2024. (Photo by ETIENNE LAURENT/AFP via Getty Images)
South African businessman Elon Musk arrives at the Tenth Breakthrough Prize Ceremony at the Academy Museum of Motion Pictures in Los Angeles, California, on April 13, 2024.
Image Credits: ETIENNE LAURENT/AFP / Getty Images

Launch of the week

Congratulations to Rocket Lab for nailing its 50th Electron rocket launch! To commemorate the massive milestone, the company posted this pretty poignant video on X that sums up just how far Rocket Lab has come.

What we’re reading

Did you know…that we posted the agenda for this year’s space programming at TechCrunch Disrupt? We are incredibly excited by this year’s lineup, which includes some of the top founders and investors operating in the space industry. Plus fireside chats with none other than Rocket Lab’s Peter Beck and Bridgit Mendler of Northwood Space. Click the link above to learn more.

This week in space history

In last week’s ‘This week in space history’ column, we detailed the flight of Sally Ride, the first American woman to go to space. This week we’re commemorating her return. On June 24, 1983, she concluded her historic trip when the Space Shuttle Challenger touched down in California.

Image Credits: NASA

Sift is building a better platform for analyzing hardware telemetry data

by with No Comment Tech

Sift founders Austin Spiegel and Karthik Gollapudi

Image Credits: Sift (opens in a new window)

Less than a year after closing its seed round, software-for-hardware startup Sift announced a $17.5 million Series A led by Google’s venture capital arm GV to scale their platform for analyzing real-time data from hardware systems.

The company is developing a platform that provides a single source of truth for telemetry data. Such data is essential for engineers to understand a machine’s performance; even tiny anomalies, if missed, can spiral into catastrophe. One timely example that Sift provides is the uncrewed Starliner test mission in 2019, which experienced a software error that sent the spacecraft into the wrong orbit entirely and led to further delays and mounting expenses in the spacecraft program. 

Such errors could be avoided with a more comprehensive, yet simplified, software stack for telemetry data, Sift suggests. As opposed to the fragmented sensor data that must be managed by entire teams, or else stitched together with ad hoc solutions, the company is offering nothing less than what it calls “a new paradigm”: a single platform that unifies hardware sensor data ingestion, storage, and review. 

Automation is one of Sift’s biggest differentiators. In the past, a customer may have manually run tests and checked dashboards to ensure hardware health, but with Sift, they can encode “rules” into the platform instead. Sift evaluates those rules against simulations, tests, and operations, and only flags an engineer for data review when it discovers an anomaly. 

“Dashboards are fundamentally the wrong solution for in-depth data analysis because there is too much noise for a human to find the signal,” Sift co-founders Austin Spiegel and Karthik Gollapudi explained in an email.

Over the next 12 months, Sift is aiming to boost every part of the software stack with artificial intelligence, from more robust anomaly detection to data review. The company is also looking to further automate parts of the compliance and regulatory review process, as these certification workflows will help engineers communicate their readiness to regulators and cut through red tape faster, Sift says.

The company already has a list of customers, including many well-known space and hardware startups, like K2 Space, Astranis, and True Anomaly. Spiegel and Gollapudi, two ex-SpaceX engineers, said that “a startup’s need for speed, flexibility and competitive advantage drives it to adopt new tech quickly, such as Sift.”  

“Building internal tools requires a dedicated headcount to create, manage, and maintain and takes years,” they added. “Startups are focused on building their business as fast as possible, so building internal tools — something they may be able to do — is not a priority. Engineering hires and priorities are focused on their product. Hiring engineers with domain expertise to build a highly scalable data storage and analytics solution is challenging.” 

Sift currently has 16 full-time employees and expects to more than double that number over the next 12 months. 

Payabli is building payment management tools for software startups

by with No Comment Tech

Contactless payment with credit card.

Image Credits: Getty Images

Joseph Phillips and William Corbera, both of whom come from entrepreneurial backgrounds, have been friends for over a decade.

Corbera co-founded RevoPay, a payments processing platform that was acquired by payments solutions firm OSG in 2022. Phillips, for his part, led the national sales team at Seamless before heading up sales at ServiceTitan, a web-based management tool for construction contractors.

In 2020, Phillips and Corbera — having worked in payments-related jobs for a number of years — decided to team up to found their own payments-focused venture called Payabli. Payabli builds the infrastructure that allows companies, specifically software companies, to embed and facilitate payments through APIs.

“Payabli builds payment acceptance and issuance solutions [and] payment operations tools,” Corbera told TechCrunch. “We make software companies payments companies by giving them payment-facilitating capabilities without the heavy lift, administrative burden and exorbitant cost of becoming a payment facilitator.”

Payabli is essentially trying to disrupt traditional payments facilitators like Stripe, Adyen and Paytrix: Companies that let customers accept electronic payments using their platforms. Payments facilitators act as middlemen between businesses and their banks, delivering the back end for payments processing.

Payabli
Image Credits: Payabli

Payabli offers the standard array of “pay-in” payment acceptance tools, including tools to let a company’s clients make recurring or scheduled payments or request invoices. But it also provides “pay-out” tools to help companies themselves pay vendors and suppliers, like virtual credit cards, physical checks and bank integrations.

Payabli’s services extend to various “payment operations” products, as well, including products designed to mitigate risk and fraud, handle disputes and compliance and facilitate underwriting.

“Payments and other fintech programs are the lowest-hanging fruit for software companies to unlock new revenue and create stickier, more valuable customer relationships,” Corbera said. “This is not only true for software companies, but any entity that coordinates money movement between payers and recipients.”

Payabli’s go-to-market approach has won approval from VCs, who’ve poured a substantial amount of capital into the startup. Payabli this week announced that it raised $20 million in a Series A funding round led by QED with participation from TTV Capital, Fika Ventures and Bling Capital, bringing the company’s total raised to $32 million at a “nine-figure” valuation. (Corbera wouldn’t reveal the exact amount.)

Payabli has around 60 customers, Corbera said, adding that revenue grew 3x over the past 12 months to “seven figures.”

“The new round of funding will be used to drive further product innovation, reinforce security and scalability, fuel new customer acquisition and empower existing software partners to integrate and activate total processing volume easier and faster,” Corbera said. “We had over 16 months of runway left when we raised, but we chose to raise opportunistically to further accelerate our growth and take on some large enterprise customers.”

Payabli, based in Miami, has 49 employees and expects to have nearly 70 by the end of the year.

Vybe app

Two Oxford PhDs are building an app to let you remix photos into memes

by with No Comment Tech

Vybe app

Image Credits: Vybe

Earlier this month, Google released a new feature with the Pixel 9 series phone to let users add the photographer in the group photo by swapping someone out and taking another photo. A new social network by a duo of Oxford PhDs is working on an app to let you add friends to a photo in a more memeable and fun way.

Vybe is a mobile app, which lets you upload a selfie and put your face in any picture or an existing template. It also swaps faces of you and your friends in a picture with multiple people. The tools also change the skin tone of the person in the template based on your skin tone so there is no mismatch between face and the rest of the body.

Users can see photos created by their friends in a feed and upload their own templates too. The startup thinks that this will nudge more people to create content on the platform. Vybe said it has more than 100,000 user-generated templates, with the number growing quickly because it requires little effort to create a template.

The company has built a social network around this to encourage people to connect more. While this is the core proposition now, the startup wants to build more photo- and video-related tools combining social, AR and generative AI.

Vybe has raised $4.75 million in seed funding in a round led by Stellation Capital with support from Scribble Ventures, Coalition Operators, Neo and Blueprint FTC. Plus, individuals such as NFL player Kelvin Beachum, early investors in Facebook and Uber, Ali Partovi and OpenAI’s Chief Product Officer Kevin Weil contributed to the round.

The company was founded by Mandela Patrick and Arnab Ghosh, who went to work at Meta and Snap, respectively. Patrick’s research focused on multi-modal AI video and image understanding. Later, he worked on core AI algorithms for Instagram Reels. Meanwhile, Ghosh researched generative models for image rendering. He worked at Snap on launching gen AI features like Bitmoji background.

The co-founders said that they wanted to build a social product that has a network effect value. They also believe that there is value in creating a one-click tool for actions — such as placing faces of you and your friends in a photo — that used to require a long time.

“When I was at Snap, I saw how little details can make a difference in adoption. While Photoshop was available for years, Snap allowed users to have different experiences with Lenses at a larger scale,” Ghosh said.

While face-swapping apps such as Reface already exist, the Vybe co-founders hope that the ease of creating templates and users remixing them will create a network.

“If you think about successful social products, they had a marquee format. For example, in Snapchat it was disappearing messages, and we think AI-powered photos with your friends is that format for us,” Patrick said.

The company is exploring brand partnerships, ads and subscription revenue streams. It is also gearing up to launch features like video and animation support in the app.

Peter Boyce II, the founder of Stellation Capital, believes that the founders have a deep level of depth and interest in AI, which will help them figure out the platform better. He thinks that the product being simple and frictionless to use will help Vybe in becoming a hit.

“When all the technology they have worked with fades into the background and what is left is a simple, easy, and fun way to engage with different folks. And that for me is ingredients for something that become mainstream,” he told TechCrunch.

BeReal, which banked on a format of nudging people to share in real time when they got a notification, faced issues when it comes to maintaining growth and generating revenue. The company eventually got acquired by the French mobile apps and games publisher Voodoo.

Vybe co-founders and investors believe that the company will succeed because it wants to create more formats and involve users in the creation process. But they emphasized that it’s important not to create behavior that creates fatigue amongst users.

The app is available for download on iOS and Android.

Orby is building AI agents for the enterprise

by with No Comment Tech

Image Credits: Getty Images

AI “agents” are generative AI models that can perform actions autonomously, like copying info from an email and pasting it into a spreadsheet, and have been hailed as productivity superchargers. That might be a bit premature, given models’ tendency to make mistakes. But at least a few founders (and analysts and investors) seem convinced that agents are the next frontier in generative AI.

Bella Liu and William Lu are two such founders. Their company, Orby AI, is building a generative AI platform that attempts to automate a range of different business workflows, including workflows that involve data entry, documents processing and forms validation.

Lots of startups offer tools to automate repetitive, monotonous back-office business processes (see Parabola, Tines, Sam Altman-backed Induced AI and Tektonic AI, to name a few). Incumbents, too, like Automation Anywhere and UiPath, have moved to embrace AI to try to maintain pace with the generative AI competition.

But Liu and Lu claim that Orby’s tech stands out for its ability to learn and act on workflows in real time and to understand the patterns and relationships within an enterprise’s unstructured data.

“Orby’s platform observes how workers do their work in order to automatically create automations for complex tasks that require some level of reasoning and understanding,” Liu, Orby’s CEO, explained. “An AI agent installed on a worker’s computer effectively watches, learns and generates automations, adapting the model as it learns more.”

With Orby, which launched out of stealth in 2023, Liu and Lu say that they sought to create AI that could understand some of the low-level decisions being made by workers and abstract those decisions away, freeing up workers to focus on headier things.

Liu previously led AI and automation efforts at IBM, including product planning and AI-related mergers and acquisitions, and was UiPath’s director of AI product management. Lu is a former Nvidia systems engineer who joined Google Cloud as an engineering lead, helping to design generative AI document and database extraction tech.

Orby’s purported secret sauce is a cloud-based generative AI model that’s fine-tuned to complete customer tasks, such as validating expense reports. The model relies partly on symbolic AI, a form of AI that leverages rules, such as mathematical theorems, to infer solutions to problems.

Orby
Orby’s generative AI observes tasks performed by people, then learns to automate these tasks.
Image Credits: Orby

Symbolic AI alone can be inflexible and slow, especially when dealing with large and complicated datasets. It needs clearly defined knowledge and context to perform well. But recent research has shown that it can be scalable when paired with traditional AI model architectures.

“For the last two years, we’ve been engineering this AI model, and have performed successful trials,” Liu said. “There are few pure-play generative AI companies attacking the enterprise head-on with something end-to-end. We are one.”

Liu says that Orby’s model can intelligently adapt to changes in workflows, like when an app’s UI gets an update, by analyzing API interactions and a worker’s browser usage. Having software monitor an employee’s every move sound like a privacy disaster waiting to happen. But Liu claims that Orby doesn’t store most customer data; it only uses certain telemetry data to improve its model, encrypting the data both in transit and at rest.

“Humans are kept completely in the feedback loop,” she added.

Orby, which recently raised $30 million in a Series A funding round co-led by New Enterprise Associates, WndrCo and Wing (sources say at a post-money valuation of $120 million), is competing in a challenging sector. Forthcoming agentic AI from generative AI powerhouses such as OpenAI and Anthropic have dampened the prospects of incumbents and smaller players alike.

Adept, a startup building AI agents technology focused on enterprise applications, is reportedly on the cusp of an acqui-hire deal with Microsoft before it manages to ship a single product. Amazon and Google have released AI agent tooling to little fanfare. Elsewhere, UiPath — despite its ramping up of generative AI initiatives in the past year — saw sales plummet in its most recent fiscal quarter.

Liu says that Orby can come out ahead by taking a systematic go-to-market approach. The company is already generating revenue from around a dozen customers, she says, and plans to put its $35 million war chest toward expanding its Mountain View-based, roughly 30-person team.

“The funds are being used to scale our go-to-market, customer support, product and technical orgs,” she said. “The enterprise market has an insatiable appetite for generative AI solutions that demonstrably improve business performance; they are just trying to figure out where to best apply the technology in the near term before they scale it across their business.”

X is building a 'dislike' button for downvoting replies

by with No Comment Tech

An illustration of a phone and the X logo.

Image Credits: Bryce Durbin/TechCrunch

Elon Musk’s X, formerly Twitter, is continuing to develop a downvoting feature that will be used to improve how replies are ranked. Although the company has not yet officially announced its plans, more recent findings indicate the downvote feature may actually resemble a “dislike” button instead of a Reddit-style downvote icon. Code references found in the X iOS app now show a button that appears as a broken heart icon next to X’s heart-shaped “like” button as well as direct references to a “downvote” feature.

As Twitter, the company tested downvoting in 2021, ahead of Elon Musk’s acquisition. At the time of the original experiment, however, Twitter had tested both upvoting and downvoting buttons across all posts. The latest tests indicate that X is only considering allowing downvotes on replies, to help showcase the better replies at the top of a long thread while moving less-liked replies further down the thread. That could prevent users from posting content designed specifically to anger people and trying to generate dislikes as a form of engagement.

Earlier this month, reverse engineer Aaron Perris, @aaronp613 on X, found references in X’s iOS app that referenced a downvote feature that appeared to be in development. He’s now found additional image files in the iOS app that indicate the button could be styled as a broken heart as well as more direct references to the feature itself.

In screenshots shared on X, Perris found that X’s app now includes several newly added references to a “downvoting” function as well as strings of text that ask the user to take action and confirm their downvote. For example, one reads “Do you want to downvote this post?,” while another simply instructs the user to “Downvote this post.”

Given the wording — which references “posts” and not just “replies” — it’s not clear if X is now considering bringing a downvote feature to all posts on the platform or only just replies.

Another user, @P4mui on X, also shared videos of the dislike button in action, including one where a user asked them to dislike their reply to the post. The user, who had enabled the dislike button using a feature flag, additionally noted that the button was only available on replies for the time being, but they weren’t sure if that would later change.

The dislike button was also reportedly spotted on an X employee’s account who had shared a video demo of a new way to expand replies. That post was quickly deleted and reposted without the dislike button in view.

Given the growing number of sightings, it seems likely that more public tests of a dislike button are underway.

This isn’t the only change X has made to its “likes” system under Musk’s ownership. More recently, X began to hide likes from public view, allowing people, as Musk put it, to like more “edgy” content and protect their image.

TigerBeetle is building database software optimized for financial transactions

by with No Comment Tech

Image Credits: Bortonia (opens in a new window) / Getty Images

After doing some consulting for Microsoft to develop protections against zero-day exploits, software engineer Joran Dirk Greef worked with Coil, a web monetization startup in San Francisco, to help build its payments infrastructure. At the time, Coil was using a traditional database to store and process transactions. But Greef had the insight that a specialized database could prove to be much more nimble — and powerful.

The idea morphed into a skunkworks project at Coil, and Greef became a staff engineer working full-time on a new database design called TigerBeetle. Two years into the project, after customers started requesting enterprise support for the database, Greef spun out TigerBeetle as a startup.

TigerBeetle’s open source database is engineered for financial online transaction processing, Greef says, capable of handling more than 8,000 debit and credit card transactions in a single query. One query for 8,000 transactions might not sound like a lot. But most general-purpose databases would require 1 to 10 queries per transaction. And more queries translates to more latency — especially if the database is hosted on a remote server somewhere.

“TigerBeetle is a financial transactions database that provides debit/credit primitives out of the box and enforces financial consistency in the database without requiring a developer to cobble together a system of record from scratch,” Greef said.

“TigerBeetle is ideal for use cases where you need to count anything of value — not necessarily money, but including money — moving from one person or place to another,” Greef said. A common application is an internal ledger for a company like Transferwise, he added, which has to keep track of lots of money moving between accounts.

Spinning out TigerBeetle was a wise decision in hindsight. TigerBeetle recently closed a $24 million Series A round led by Spark Capital’s Natalie Vais with participation from Amplify Partners and Coil, bringing its total raised to more than $30 million. A source familiar with the matter tells TechCrunch that TigerBeetle is valued at around $100 million post-money.

“We had planned to raise later in the year,” Greef said. “However, after a surge in community growth at the beginning of 2024, and growing commercial interest, we decided to bring the raise forward to invest in engineering, go-to-market and TigerBeetle’s cloud platform, which is under development.”

TigerBeetle, which only has eight employees at present and plans to double the size of its team by 2025, provides its database technology in the form of a managed service. Greef claims that TigerBeetle has had paying customers “since day one” and that the TigerBeetle community — folks using or contributing to the open source release — has grown over 200% year-over-year.

Vais told TechCrunch that TigerBeetle is one of the more exciting database projects that she’s seen recently.

“TigerBeetle rethinks every component from the ground up to handle modern transactional workloads,” she said. “In a world where everything is becoming more online and more transactional, there’s a huge opportunity for TigerBeetle to become a foundational piece of infrastructure for modern systems of record.”

TigerBeetle’s managed service is currently available by invitation only, and the database reached its first production release just in March. But Greef says that growth — in particular acquiring new customers — will be the focus for the foreseeable future.

“TigerBeetle’s use cases extend beyond fintech,” he continued. “Think usage-based billing with real-time spend caps, gaming live ops and energy smart meters, as well as instant payments, core banking, brokering, inventory, shopping carts, trucking and shipping, warehousing, crowdfunding, voting and reservation systems.”

Tezi is building an AI agent for hiring managers

by with No Comment Tech

HR concept of robot hand selecting a candidate from a bunch of rectangles.

Image Credits: fatido / Getty Images

AI agents are all the rage right now, and Tezi, an early-stage startup, is working on one to help HR teams find the perfect candidates for a job opening. The startup claims this bot will sift through resumes to find the ones that match the hiring criteria, find time on the recruiter’s calendar to set up an interview and send out the email to the candidate.

Today the company announced a $9 million seed to help fuel its journey to generally available product.

For now, they are rolling out the alpha product with a handful of design customers just this week, but that’s the vision, according to CEO and co-founder Raghavendra Prabhu. He acknowledged that HR has been using automated resume screening for some time, but Tezi saw an opportunity with the new generation of large language models (LLMs ) to build a more sophisticated recruiting tool for HR.

“I think it’s the combination of reasoning and natural language that we felt gave us an option to build something very, very different from what’s historically been done by software in this space,” he said.

His co-founder and COO, Jason James, says that existing tools don’t go far enough in his view. “Let’s say you get a thousand applications for a job. AI or ML or algorithms in the past would be good at saying these resumes are very good,” he said. “But a human still needs to send emails and schedule interviews and all of that. And what’s possible now is an end-to-end workflow, not just basic ranking.”

The founders acknowledged that at this stage, humans need to stay engaged in the process and the hope is that it will be fully automated as models improve. What’s more, the pool of candidates that emerge from any job search is going to be dependent on the quality of the prompts and job descriptions.

Tezi interface showing HR bot interacting with a recruiter.
Image Credits: Tezi

While they understand that automation can lead to bias, they are working on mitigating that to the extent possible. From their perspective, they are taking whatever inputs come from the hiring manager and assessing that against the resumes in an objective manner. They can’t control what the inputs look like, but they say they are trying to minimize bias on their end.

“If you’re assuming that there’s bias coming in from the employer, we at this time aren’t going to be excellent at preventing that. What we will be doing on our side is protecting against us adding any sort of bias into the mix through algorithmic means,” James said. They are avoiding looking at historic hiring patterns. They want the models to match by skills and other criteria set by the hiring manager.

They have trained their models on 250 million profiles that they have licensed from data providers and have been working with OpenAI and Anthropic models so far and tuning them to their hiring requirements. 

The company is just starting. It launched at the beginning of this year. They are beginning work with 15-20 design customers, and the hope is that they will work out all the kinks and get to a wider beta distribution later this year.

The $9 million seed was led by 8VC and Audacious Ventures with participation from Liquid 2, Afore, PrimeSet, South Park Commons and industry angels.