a16z speedrun SR005 companies

Agent Astra builds an AI-native freight forwarding platform that owns and operates the logistics businesses running on it. The platform gives freight operators tools to move containers, track shipments, and manage complexity with a fraction of the headcount traditional brokers and forwarders require. It targets the full service range of a global logistics provider while compressing the workforce needed to deliver it.

Ambiguous AI builds a unified productivity workspace where AI agents and human teammates work side by side in the same apps on the same data. The suite covers 17 tools, including Docs, Mail, Chat, Sheets, CRM, and Calendar, each designed so every action a person can take an agent can also take natively. It replaces the hand-off friction of bolting AI onto existing tools by treating agents as first-class coworkers from the ground up.

Anchr builds a workforce of AI agents for food distributors and other physical operations businesses, automating the manual, repetitive work that runs across order intake, procurement, inventory management, and vendor coordination. Each agent plugs into the distributor's existing systems and handles discrete operational tasks, replacing the human hours spent routing orders, matching invoices, chasing vendors, and resolving unit discrepancies. The product targets the operational core of distribution businesses, where errors and delays compound across every transaction.

AndThen builds a voice-first interactive entertainment platform where users play through short, episodic experiences, such as game shows, mysteries, and character-driven stories, by speaking aloud. Each format is designed to be replayable and structured around real goals, replacing passive audio content with conversation-driven gameplay. Creators, writers, celebrities, and brands publish formats on the platform, with user-generated publishing on the roadmap.

Argu AI builds a natural-language video intelligence platform that turns existing CCTV cameras into configurable monitoring agents. Security and operations teams at ports, utilities, government facilities, and commercial venues describe what they want to detect in plain text, and the system creates alerts, searches, and analytics across thousands of cameras without writing code or retraining models. The product replaces lengthy procurement cycles, custom model development, and fixed preset-based surveillance systems.

Arise Labs builds an AI investment analyst that ingests unstructured data, constructs auditable investment theses, and executes decisions with the rigor of an institutional analyst. It targets portfolio managers and investors, replacing the manual research and decisioning work typically done by human analyst teams. The system is designed to run continuously, delivering institutional-grade coverage at any scale.

Artifact AI builds an AI agent platform for accounting and professional services firms, automating the manual work of bookkeeping, ledger posting, reconciliation, and financial close. It replaces human-hours spent on transaction review and reporting by deploying purpose-built agents that score entries, generate audit trails, and surface real-time client metrics. The target customer is the accounting firm looking to increase capacity without adding headcount.

Atrios turns well-connected individuals into a paid referral network for B2B software companies. Tastemakers share personalized links to vetted SaaS products and earn flat fees per booked meeting and larger payouts when a deal closes. The platform replaces cold outbound sales with warm introductions from people who already have relevant relationships.

Avenir AI builds AI agents that automate employee benefits management for self-insured employers and benefits brokers. The product analyzes claims, utilization, and regulatory data to optimize plans and vendor decisions, replacing the manual work brokers and HR teams currently do by sifting through large, fragmented datasets. It targets a market where legacy systems have made benefits administration both costly for employers and opaque for employees.

Axon builds brain-inspired AI models for high-frequency financial trading, where speed of prediction is the competitive edge. Its architecture anticipates market moves before full data arrives, cutting latency at both the model and infrastructure layers. The firm targets trading workflows that conventional low-latency systems and standard deep learning models handle separately, combining both into a single research-led stack.

Bead AI builds AI agents for internal audit teams, starting with Sarbanes-Oxley compliance testing. The agents collect evidence from source systems, validate submitted files, execute existing testing plans step by step, and produce marked-up working papers, replacing the manual document chasing, screenshot comparisons, and spreadsheet population that consume most of an auditor's time. The target customer is any US-listed company running an internal SOX audit program.

bins builds a peer-to-peer secondhand marketplace designed to replace the scroll-and-search model of platforms like Depop and Poshmark. Each morning, users receive a curated, finite set of resale clothing collections, called bins, rather than an open-ended catalog. The product targets style-conscious shoppers and casual sellers who want the tactile surprise of thrift-store browsing without leaving their phone.

BotBot gives non-technical teams, brand, design, marketing, and product, direct control over voice AI behavior without engineering tickets. Teams edit tone, pacing, persona, and upsell timing in plain English, run A/B tests against live customer conversations, and read the results in metrics like ticket size, CSAT, and containment. The product replaces static, launch-day configuration that otherwise runs unchanged on every call until someone files a ticket to change it.

Brief builds a product-context layer for AI coding agents, giving them access to the decisions, requirements, and institutional knowledge that live across tools like Linear, Jira, Notion, Slack, and GitHub. Product and engineering teams use it to keep agents aligned with existing product intent rather than letting them build against stale or missing context. It replaces the manual work of re-explaining decisions to every agent and the rework that follows when agents build without that grounding.

Color builds an engine that brings AI to life by fusing a AAA game engine with multimodal machine learning models. The engine powers Color's first consumer app, Otherhalf.ai, an AI companion designed for human presence and connection rather than information retrieval. This replaces the text-only, question-and-answer model that defines most AI products today.

Creed builds a Christian AI companion app for everyday believers, pastors, and ministry leaders. Users can ask faith questions by text or voice, with every response grounded in Scripture alongside readable verse references. The app replaces ad-hoc Google searches and generic devotional content with guided daily prayer, journaling, Bible reading, and plain-language Scripture explanation built around Christian values.

Dex builds an AI recruiter and career coach for software engineers, replacing the standard CV-and-job-board process with a conversational intake that maps each engineer's ambitions, motivations, and deal-breakers. It uses that profile to match engineers with tech companies and introduce them directly to hiring managers. On the company side, Dex surfaces pre-qualified, often passive candidates who are matched on fit rather than keyword overlap.

Dispatcher builds an aerial intelligence platform that converts commercial drones into autonomous agents capable of making real-time decisions without a human operator in the loop. Security teams, construction sites, agricultural operations, and heavy industry use it for automated perimeter patrols, incident response, and asset monitoring, replacing manual drone piloting and after-the-fact data review. Operators issue commands in natural language and receive live alerts, heatmaps, and reports through app, SMS, and dashboards.

doublespeed builds autonomous social media accounts that post, engage, and grow audiences on TikTok and Instagram around the clock. Brands and marketers use it to drive organic attention at scale across any demographic niche, replacing the human labor of creator sourcing, content production, and community engagement. Each agent learns from performance data over time, making future content more effective.

EverCurrent builds an AI-native platform of record for hardware engineering teams, connecting bills of materials, ECOs, handoffs, and team decisions into a single view. It automatically documents changes across tools and teams, ties each decision to its broader context, and surfaces dependency risks before they become costly. The product replaces fragmented, manual tracking of hardware changes that leaves teams blind to downstream impacts.

Ezra AI Labs builds a voice AI interviewer for corporate recruiting teams, replacing the initial human screening call. Recruiters send candidates a structured voice interview conducted by the AI, which captures spoken responses that resume reviews and text-based screens cannot surface. The product targets the gap created by AI-generated resumes and coached answers by evaluating candidates on nuance and authenticity at scale.

Fearn handles patent prosecution for startups, replacing the slow, expensive billing model of traditional Big Law firms. Former Big Law patent attorneys review every filing, and AI handles the drafting work that previously drove up hours and cost. Startups get portfolio-grade patents on days-long turnarounds instead of quarters-long ones.

First Voyage builds a mobile self-care app centered on Momo, an AI-powered mythological creature that turns daily wellness habits into a game. Users complete small quests around sleep, journaling, and personal routines, earning in-app coins to customize Momo and its world. The product replaces passive habit trackers with an immersive, character-driven loop designed to keep people returning to their routines daily.

Forgepoint builds AI agents for the manufacturing back office, starting with Frank, an AI inside sales rep that handles the full RFQ-to-quote workflow for job shops and industrial manufacturers. Frank qualifies incoming requests, prices parts against the shop's cost model, sends quotes for human approval, follows up with customers, and logs every action to the CRM and ERP. It replaces the manual quoting work done by sales reps, estimators, and engineers across injection molding, metal fabrication, CNC machining, and adjacent trades.

General Magic builds a suite of insurance-specific AI agents that handle the operational workload inside insurance brokerages: underwriting questions, customer servicing, and compliance review. Edward answers appetite and eligibility questions by citing carrier manuals directly, replacing manual document searches. Ava services policyholders around the clock over email, text, and voice, handing off to a licensed broker with full context when needed, while Peter reviews every conversation for quality and compliance.

Jigsaw builds post-training data for recursive self-improvement, selling to frontier AI labs working toward advanced AI systems. The team converts AI research and development into verifiable, gradable tasks, then supplies the resulting artifacts as post-training data, replacing the ad-hoc and hard-to-evaluate methods labs currently use to generate training signal for self-improving models.

Jooba runs end-to-end recruiting for high-growth teams, replacing the work done by traditional staffing agencies and in-house recruiters. AI agents handle sourcing, outreach, scheduling, and reporting, while hiring managers review shortlisted candidates and schedule interviews directly inside Slack. The product targets the top of the hiring funnel, delivering vetted candidates without requiring teams to manage the search themselves.

Kanu AI builds an enterprise AI platform that converts repeatable internal workflows into purpose-built software driven by a firm's own data, documents, and institutional knowledge. It targets data-heavy financial services verticals, including commercial real estate, private equity, insurance, and financial services, replacing manual research, analysis, and reporting work that teams currently do by hand. The platform connects disparate data sources, lets users query them in plain language, and delivers finished outputs in the formats teams already use.

Kintow automates back-office operations for restaurants, connecting fragmented point solutions, such as POS, scheduling, invoicing, and banking, into a single system that tracks food costs, labor, and cash flow. It replaces manual forecasting and ordering by anticipating demand and sending vendor orders automatically, freeing managers from spreadsheet work. The target customer is multi-location restaurant operators whose general managers currently juggle disconnected tools to run daily operations.

Lekondo builds a social platform centered on fashion, giving consumers a dedicated space to discover, discuss, and engage with style outside of general-purpose social networks. The product targets fashion enthusiasts and replaces scattered fashion conversation that currently happens across mainstream social media.

Limy AI builds an AI Search platform for brand marketers, showing them how AI agents discover, cite, and recommend their brand across AI-powered search experiences. It replaces guesswork about AI-driven discovery with attribution, analytics, and visibility tools that connect agent behavior on a site to downstream customer actions. Marketers use it to find gaps in their AI Search presence and measure the traffic and revenue that result.

Loan Labs builds AI workers for mortgage companies, starting with LOIS, an AI loan processing assistant. LOIS collects and organizes borrower documents, extracts structured data, flags inconsistencies, and automates borrower follow-up, replacing the manual work done today by loan processors and underwriters across fragmented email, CRM, and loan origination systems.

Logical Health builds an AI-driven member engagement platform for health insurers and healthcare organizations. It replaces one-way outreach, such as unanswered calls and unopened emails, with two-way text conversations that identify the specific barrier keeping a member from acting and resolve it in the same thread. Underneath the conversational layer sits a rules engine that converts insurance policy language into precise, machine-readable logic to answer coverage questions accurately.

Luvu builds a human data foundation that backs frontier AI systems, giving AI agents grounded, people-centered context to draw on. The product targets consumer-facing AI applications where understanding human behavior and preference is the core input. It replaces ad hoc data curation and hand-labeled behavioral datasets that teams otherwise assemble manually.

Margin builds AI-native credit card infrastructure that lets consumer brands offer dynamically personalized rewards, replacing the static, one-size-fits-all loyalty programs that traditional card issuers run. The platform analyzes individual spending behavior in real time and adjusts rewards to match each cardholder's preferences. Margin is extending the model beyond credit cards toward a broader suite of personalized consumer credit products.

Metal builds an end-to-end fundraising platform for founders, combining industry data with company-specific context to run the workflows that make up a capital raise. It replaces the spreadsheets, manual research, and disconnected outreach tools founders typically stitch together when going out to investors.

Moescape AI builds an all-in-one creative platform for anime fans, letting users design original characters and then generate images, videos, music, and voice output from them in a single environment. Fans can also chat with their characters through an LLM-powered roleplay system. The platform replaces a fragmented workflow that previously required separate tools for art generation, voice synthesis, video creation, and interactive fiction.

Moona Health builds a virtual network of CBT-I therapists who treat chronic insomnia online, covered by insurance. Patients work one-on-one with licensed clinicians through a short course of cognitive behavioral therapy for insomnia, replacing sleep medication and unguided wearable apps. Behind the clinicians, an AI platform handles insurance claims, scheduling, and sleep data analysis to give therapists personalized session guidance.

Nexxa.ai builds agentic AI software for heavy industrial enterprises, automating critical engineering and operational workflows that today depend on institutional knowledge held by individual workers. Its AI agents write directly into customers' existing systems of record, replacing the slow, manual decision cycles typical of generic ERP platforms and consulting engagements. The result is software purpose-built for a single operation, maintained continuously by AI, and owned outright by the customer.

Ocoder builds an AI-powered development environment for designers and product managers who need to work directly with real codebases. It lets non-technical team members prototype and build new features without engineering support, then share results with engineers and stakeholders via shareable links. The product replaces back-and-forth communication cycles where non-technicals previously had to relay ideas through developers to see them realized.

Oleum builds AI tools that let business users query, explore, and act on data without relying on specialized analysts or engineering pipelines. Built by a team of former quantitative researchers, the platform targets data-oriented businesses that need faster, more accessible insight from their existing data. It replaces manual data wrangling, ad hoc analyst requests, and bespoke internal tooling.

Pen.dev builds an agentic design canvas that lives inside a codebase, letting engineers design UI without handing off to a separate design tool or team. Agents create, refine, and iterate on layouts directly on the canvas, replacing the back-and-forth between Figma files and code editors. It connects to Cursor, Claude Code, VSCode, and OpenAI Codex through MCP, and stores designs in an open format alongside the code.

Prepp builds AI-powered modules for students and educators, handling tasks that traditionally require manual effort: generating course materials, grading assignments, tracking attendance, and driving student engagement. The product targets both the administrative burden on teachers and the content delivery experience for learners, replacing disconnected, labor-intensive workflows with a unified toolset.

Presia builds AI agents for management consultants, automating the manual, process-heavy work that consumes time consultants would otherwise spend on analysis and client problem-solving. The product targets consulting firms and enterprise clients, replacing the routine document, research, and synthesis tasks that make up the bulk of a typical engagement's labor.

Rork lets non-technical founders build, install, and publish native iOS, Android, and web apps by chatting with AI, replacing the need to write code or hire mobile developers. Users describe what they want and Rork generates production-ready Swift or Kotlin apps that can be submitted directly to the App Store or Google Play. The product targets aspiring app founders who want to go from idea to a live, monetizable app without an engineering background.

Sans Strings Studio produces transmedia entertainment, building stories and IP across games, film, social, and live events from a single shared asset pipeline. Its F.E.L.T. system lets IP holders generate both interactive and linear content simultaneously, replacing the separate production workflows each format normally demands. The studio targets the gap between rising content demand and rising production cost, using real-time technology to scale output while keeping human creative authorship intact.

Shockwave builds an AI-powered daily games platform targeting the casual gaming market. The platform delivers a growing roster of daily games across multiple genres, competing in the space that hits like Wordle opened up. It replaces one-off game downloads and ad-hoc play with a single destination for recurring daily engagement.

Sonatic builds a desktop AI assistant for knowledge workers that watches how a user works across their tools, learns their patterns, and then executes routine tasks on their behalf without manual commands. It targets the repetitive coordination and execution work that currently consumes workers' time in apps like email, project management, and other workplace software. The product replaces the manual, step-by-step task completion that workers would otherwise do themselves.

SoulLink builds AI companion apps that give users a persistent, cinematic virtual relationship. The product pairs 3D character rendering drawn from AAA game production with an emotional AI layer that retains conversational history across sessions. It targets consumers seeking ongoing social connection, replacing the flat, stateless chat experience of earlier virtual companion tools.

Sounding builds voice-first AI wellness apps for consumers, targeting the adherence problem: getting people to follow through on health behaviors over time. The product replaces passive content and manual check-ins with an interactive voice experience that guides users through real-world wellness routines.

Sourcerer acts as an AI-native, full-stack distributor for companies buying physical goods across global supply chains. Its autonomous agents handle the entire procurement cycle: vetting factories, negotiating prices, arranging freight, and offering trade credit, then issuing a single consolidated quote. The product replaces the manual RFQ process, broker relationships, and spreadsheet-and-email workflows that traditional distributors rely on.

Syzygy builds ultra-compressed, task-specific AI models and the local inference stack to run them on Apple Silicon, targeting enterprises and developers who want to own their AI layer instead of routing queries through third-party API vendors. Its Mach 1 Additive architecture runs a 35B mixture-of-experts model at 1.7 bits per weight, retaining 96.3% of full-precision benchmark scores while fitting on a 16 GB Mac. The product bundle spans a desktop coworking app, a local inference engine, and a TypeScript agents SDK, replacing cloud inference calls with on-device, proprietary model deployments.

Taya builds a jewelry-form-factor AI wearable for everyday consumers who want to capture and revisit fleeting thoughts without reaching for a phone. The product, a stainless steel pendant on a 14K gold-plated chain, lets wearers press a button, speak naturally, and retrieve those moments later through a companion app. It replaces the habit of typing notes mid-task or losing ideas entirely when attention is already divided.

URSA Mining builds an autonomous operating system for mining equipment, converting existing excavators and earth-movers into self-directing robots without requiring new hardware purchases. The system targets mine operators contending with a retiring skilled workforce, decade-long development timelines, and production costs that consume the majority of operating budgets. It replaces human-operated and manually supervised earth-moving with software-driven autonomy, and tracks material movement from pit extraction to final destination.

Variant builds a social platform that converts text, memes, and doodles into shareable, remixable mini-games. It targets creators and community builders who want to turn everyday content, jokes, fan fiction, or casual ideas into interactive experiences without any game-development work. The platform replaces static posting with playable, living content that communities can remix, battle over, and publish across TikTok, Discord, and Reddit.

Vega builds a relationship intelligence app that turns user-written notes about friends into curated social groups, called orbits, matched by mutual trust and shared experiences. Users trade notes on the people in their lives to receive AI-generated introductions and context about their own network. The product replaces the manual, hit-or-miss work of figuring out who in your extended circle is worth knowing better.

YKB builds a social prediction app for sports fans, letting them log takes, compete with friends, and hold each other accountable when outcomes land. An AI agent generates interactive prompts called Receipts whenever news breaks, replacing the informal, forgettable group-chat argument with a structured, timestamped record of who said what. Starting with NFL, the product targets the broader culture fan who wants bragging rights, not a betting account.

Zero Billion runs a consumer research lab that ships AI-powered consumer apps, observes how people use them at scale, and sells the resulting behavioral data to model developers. The marketplace pays contributors directly via Stripe for audio, video, screen recordings, calls, and other everyday human data, on a consent-first, revocable basis. The studio arm replaces traditional media production pipelines by letting anyone generate short-form TV-style shows in under an hour for a fraction of conventional production cost.

Zingroll operates a streaming platform for AI-generated movies and series, built for creators who produce shows entirely with generative AI tools. Viewers watch original AI-made content from a global pool of creators, replacing the traditional studio and production pipeline that previously stood between a creative person and a distributed audience.