David J Phillips
@davj • 21,033 subscribers
Founder & CEO @fondocom – accounting & tax on autopilot for 2,000+ startups. Angel: Flexport, LiquidDeath, Rippling, 100+. Host: @thestartpod
Shorts
Videos
Introducing FREE TaxPass from Fondo.com We've helped thousands of startups like Legora, Cursor, ElevenLabs, and PostHog get their corporate taxes filed Startups shouldn’t have to juggle bookkeeping and tax filings with different providers Starting today, when you sign up for monthly bookkeeping, your first year of corporate taxes are filed for free (save $1,950+): • Delaware Franchise Tax • Federal Corporate Tax (Form 1120) • 1 State Corporate Tax filing We’ll also handle Form 5472 if you’re an international founder. All filed by our CPA-led team. If you’re a Delaware C-Corp and haven’t filed yet:
David J Phillips401,152 views • 7 months ago
Somewhere, someone already has the thing you're looking for. A dataset, a patent, a capability, a customer, an answer. Most of the time neither of you ever finds out. That's the problem Ryaan Aqid built Quirk around, and he first recognized it in Bangladesh, watching workers earning $50–100 a month while Upwork listed work paying ten times more. Nothing separated them except information. The opportunity already existed; the connection never happened. It runs through everything. Institutions sit on datasets they'll never open, companies shelve technology somebody else is desperate for, and entire markets fail to form because two people who'd have built something never met. Ryaan calls it the largest economy nobody can see: the things that were never made. Quirk builds infrastructure to dissolve that asymmetry, with agents that do the looking so neither side has to move first. Holding something you won't use? It works out who it's worth something to. Stuck? It finds whoever got past this a year ago, without you having to describe the problem. "Agents that find deals neither side knew existed." 🎙️ Ryaan Aqid, Founder & CEO, Quirk on Fondo.com START 01:03 Getting the buyers to say exactly what they needed 01:31 Building distribution through government relationships in Bangladesh 02:18 Why he thinks the average person's information gets sold through intermediaries 02:40 The high school nonprofit in Bangladesh 03:08 Realizing freelance platforms could 10x a worker's income 03:25 Moving people up Maslow's hierarchy of needs 04:12 Leaving Cornell two days after classes started 05:08 What's next and where to follow along Check it out at
David J Phillips39,045 views • 1 month ago
Voice AI can pass a Turing test. For about a minute. That's a generated clip, though. Have a human actually talk back and the number collapses to six or seven seconds, roughly where generated voice sat three years ago. One reason: real conversation isn't turn-based. Around 20% of the time more than one person is speaking, and laughter drives a lot of that overlap, since you laugh at a joke while it's still being told. We also adjust our pacing toward whoever we're talking to without noticing we're doing it. Voice models struggle with all of this. Full-duplex voice, where a model listens and speaks at the same time, is still extremely early. So an agent can know your joke is funny and still have to wait until you've finished before it laughs, by which point the timing has killed it. Miso Labs CEO & Co-Founder, Aoden Teo, describes a second consequence: agents get pushed toward almost "psychotically emotive" behavior. If they can only talk once you've stopped, they need some other way to show they were listening. You finish your sentence, and the thing goes "Hmm?". You've heard it. Underneath that sits an architecture problem. Voice models have to respond fast, which constrains how large they can be, and fast means something different here than it does in text. Working with an LLM like Claude, you care how quickly it finishes your code, more than how quickly it starts. Voice inverts that. Nobody needs 10 hours of audio generated in two seconds, because nobody can listen to 10 hours of audio in two seconds; what matters is reaction time. Most architectural decisions trade latency against throughput, and Aoden expects voice to keep moving away from LLM-style designs toward ones built around very low latency. Miso Labs is already pushing on it. Miso-1 got 3,000+ stars on GitHub and 5 million views on X, and they record data in their own LA studio because the internet doesn't contain every kind of audio a voice model might need. Nobody has released a podcast of someone reading millions and millions of email addresses, and people still want voice models that can read email addresses aloud, so teams end up generating some very strange training data themselves. The clip isn't the hard part. The hard part starts when you talk back. "The most emotive foundation models for voice" 🎙️Aoden Teo, CEO & Co-Founder, Miso Labs on Fondo.com START 1:03 Miso-1: 3K+ GitHub stars + 5M X views 1:59 Why emotiveness matters for games, UGC + interactive products 3:06 Measuring progress in voice AI with longer Turing tests 4:01 Why interactive conversation is harder than generating convincing clips 5:08 Full-duplex voice, interruptions + why laughter matters 6:04 Latency vs. throughput - and why voice differs from LLMs 7:09 Miso's LA recording studio + the challenge of voice training data 9:02 Talking teddy bears, UGC, anime + unexpected voice AI use cases 10:19 From serious chess player to math obsession to building Miso Labs 12:11 The surprise YC interview
David J Phillips36,612 views • 1 month ago
AI shouldn't replace therapists. It should help them spend more time being therapists. Not because AI won't get smarter Because good therapy isn't just information It's trust It's compassion It's the human element. Most therapists spend only about half their time doing therapy The other half? Running the business Notes. Paperwork. Marketing. Insurance claims. Klarify automates the work around the session so therapists can focus on the work only they can do Today, more than 7,000 therapists use the platform The result is clear: they spend less time running a business, and more time doing the work only they can do 🎙️Moody Abdul, CEO & Co-Founder Klarify (YC P26) on Fondo.com START Huge thanks to Rippling for making this episode possible!
David J Phillips88,781 views • 4 months ago
The internet was the training set for intelligence Nobody has built the equivalent for the physical world Previously, Lucas Ngoo co-founded Carousell, scaled it $1B+ Now at CortexAI he's collecting the data robotics labs need to train foundation models Cameras, VR headsets, glasses on factory workers, retail workers, everyday people. Recording real-world manipulation work Not building the robot. Not training the model. Collecting what goes in 2026: scale data 2027+: start rolling robots out Initially 90% teleoperation, 10% autonomy While humans step in, the system keeps learning That's the flywheel toward full autonomy 🎙️ Lucas Ngoo, Co-Founder & CEO, CortexAI on Fondo.com Start (full ep in comments)
David J Phillips110,268 views • 7 months ago
Speed to lead sounds like a sales tactic It's actually the whole game Ruben Harris knows this better than anyone He built a 300-person call center across 3 countries to reach 1M people a month ...and saw exactly where the ceiling was The problem isn't effort. It's physics. A working-class student fills out a form at 10 PM. Nobody's staffed at 10 PM So 90% of leads go cold Not because they weren't interested Because nobody got back to them fast enough Outrival's AI agents respond in under a minute They call, text, email, and follow up One client made $ 10M in new revenue from leads they already had The best outbound AI isn't replacing humans It's doing the work humans never had time to do 🎙️ Ruben Harris, CEO & Co-Founder, OutRival on Fondo.com START Huge thanks to Rippling for making this episode possible!
David J Phillips80,212 views • 5 months ago
In many cases, the best AI companies aren't replacing software. They're giving it a login. Every software company for years has made the same pitch to manufacturers: clean APIs, migrate your stack, rip out your ERP It rarely works well. The ERP is the company brain. Years of data. You can't rip it out any more than you can rip out someone's memory nikolas & Co-Founder lukas figured out the obvious thing that everyone missed Stop trying to replace the software. Give an AI employee a login instead. Walter signs into SAP, Microsoft Dynamics 365, and Oracle the way a new hire would Ready to go from day 1. Works out of Teams and email like the rest of the team. Reads orders. Enters them. Places supplier POs. Catches pricing errors before they ship... Not another dashboard. Not another integration project. Not another rip and replace. A purchase order that used to take 15 minutes now takes a few seconds. The insight came from a $ 100M+ manufacturer who had a full-time employee doing nothing but typing purchase orders into an ERP Not because the work was valuable. Because the software demanded it. With Walter, your software stays. The manual work disappears. 🎙️ nikolas, CEO & Co-Founder of Walter on Fondo.com START Huge thanks to Corgi for making this episode possible!
David J Phillips59,091 views • 4 months ago
“I don't believe Claude Code will exist in its current form in six months” Benۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗ☁️, CEO & Co-Founder Freestyle, thinks coding agents are moving from local machines to the cloud, where a single task could get the attention of 20+ agents at once. Each gets a complete copy of your stack, production environment included. Each can spend a week testing and refining its approach. You compare the results and take the best one forward. The economics aren’t there yet. Ben expects cost per task to fall another 99% over the next four years, making that level of parallel work practical. Freestyle builds the computers for it: full Linux VMs for tasks that run for hours, days, or weeks. Clone a running machine, memory included, and let agents pursue different approaches from the same starting point. Pause and resume with their state intact. Inside Freestyle, Ben already gives agents a week to find improvements to its VM technology. Roughly 700 tests and 90 metrics help the team judge whether the work made things better. He calls this goal engineering: define the outcome clearly, give agents time to work toward it, and measure whether they’re making progress. "Full Linux VMs for AI Agents" Built for complex tasks that run for hours, days, or weeks 🎙️ Benۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗۗ☁️, CEO & Co-Founder, Freestyle on Fondo.com START 00:57 Why coding agents could move from local to the cloud 02:22 From harness engineering to goal engineering 03:08 Giving agents a week to improve measurable results 04:13 Why defining the problem becomes the bottleneck 04:48 How early access to o1 changed Freestyle’s direction 05:51 Why frustrated sandbox users revealed a bigger opportunity 07:28 Giving agents a computer instead of building custom tools 08:12 Getting into YC 11:17 Snapshotting VMs and running agents in parallel 12:24 The economics of letting multiple agents attempt every task 13:26 How GPU supply could drive down agent costs 15:09 Try Freestyle and follow Ben
David J Phillips14,633 views • 25 days ago
Most companies don't realize they have a documentation problem until everyone already depends on it. Customers use it. New hires use it. Engineers use it. And when documentation falls out of date, the whole system starts working against itself People stop trusting what they're reading Teams lose context And nobody can fix it, because the problem is everywhere at once The longer it goes unfixed, the harder it is to untangle That's the opportunity Manicule saw They're building an AI-native DevRel company for developer tools - owning documentation, technical content, GEO, and distribution across social channels. The premise isn't that AI should replace expertise. It's that expertise should scale. Their AI agents audit and test at scale. Their humans own the architecture, the writing, and the creative direction. Every review improves the system Every project creates more context. Every iteration raises the bar What started as a highly manual business - helping developer companies create better technical content - has become a scalable AI-native operation Today, Manicule works with some of the fastest-growing developer tool companies - including Supermemory, Greptile, and Reducto - has scaled fast over the last few months, and has more demand than it can take on. Not because they publish more content. Because they help developer companies create content developers actually trust. 🎙️ Naman Bansal & Shreyans Jain , Co-founders Manicule (YC P26) on Fondo.com START Huge thanks to Numeral for making this episode possible!
David J Phillips48,703 views • 4 months ago
Most cancer drugs target proteins FinalDose thinks the real opportunity is one layer deeper: DNA Jeff Liu and his team are building a programmable drug platform that uses genetic mutations to identify and selectively destroy cancer cells Cells without those mutations are not targeted Instead of creating a completely new drug for every cancer type, the platform is designed as a reusable system with different genetic instructions One chassis. Different disease targets. The broader bet: DNA is more deterministic than proteins, which could unlock therapeutic approaches traditional drugs struggle to reach Search. Destroy. "A programmable cell elimination platform" Jeff Liu, CEO & Co-Founder, FinalDose ( on Fondo.com START Big thanks to Numeral for making this episode possible!
David J Phillips46,679 views • 4 months ago
A production issue shouldn't require a detective Yet that's how most observability works today Most observability tools just dump alerts. Duplicates, no context, and you still fix it yourself Nicolò Magnante (YC P26) and his cofounder Arseniy Shishaev (YC P26) - who spent years building part of Datadog's metrics product - think that's backwards So they built Superlog (YC P26): Observability that's meant never to be opened A wizard scans your repo, installs proper OpenTelemetry, and runs daily to keep up as you ship When something breaks, it groups the errors into one incident, investigates with full context - logs, traces, recent deploys, past Slack threads - and drops a single mergeable PR in Slack. Merge it Ignore it Or open it in Claude Code and tweak it Vendor-neutral, so you keep every log, trace, and metric - even if you leave Not another notification. A fix. Nicolò Magnante (YC P26) CEO & Co-Founder Superlog (YC P26) on Fondo.com START Huge thanks to Eragon for making this episode possible!
David J Phillips40,983 views • 4 months ago
Your AI agent makes a mistake. Who pays for it? Not your E&O policy That assumes a human made the error Not your cyber policy That assumes an attacker breached the system Autonomous AI agents are explicitly excluded from both Ines Boutemadja discovered this while building AI agents for enterprises Then procurement started asking for proof the AI was actually covered Existing policies couldn’t provide it. So she built Klaimee: Insurance for AI Agents Built for companies deploying AI in healthcare, fintech, and voice & more Her bet on the future: AI liability is inevitable 🎙️ Ines Boutemadja (Ines Boutemadja), CEO & Co-Founder Klaimee on Fondo.com START Big thanks to Mastra for making this episode possible!
David J Phillips42,078 views • 5 months ago
Today, I’m thrilled to share that has raised a $7M seed round to accelerate our mission of putting accounting on autopilot for startups. With Fondo, startups don’t need to worry about tax penalties, financial statement errors, or missing out on up to $500k in tax credits. Onboard in minutes then set & forget — taxes are filed on time, timely and accurate financial reporting is delivered, and cash back tax credits from the IRS (up to $500k per year) are claimed and received. Over 1,000 startups trust us to put their bookkeeping, taxes, and tax credits on autopilot and starting today any startup can try Fondo for $1: We are very grateful for our supporters in this round. The round is led by マネーフォワード (Sandy Kimura, 辻庸介/Yosuke Tsuji@マネーフォワードCEO) with participation from Y Combinator (Gustaf Alströmer, Michael Seibel, Dalton Caldwell, Arnav Sahu), @foolventures (Maggie Dorn, Brendan Mathews, Abigail Malin), Next Coast Ventures (Mike Smerklo, Michael Maloney, Kaitlyn DeBernardo), twenty two ventures (Katey Caldwell), a16z scout fund (Theo - t3.gg, Warren Shaeffer), Index Ventures scout fund (Nitzan Shapira), Costanoa Ventures scout fund (Todd Johnson), 1121.vc (Mike Rosengarten), Rho (Everett Cook), Capchase debt (Miguel Fernandez) and angels including Craig J. Lewis, Chris Bakke, james hawkins, Josh Cohenzadeh, Greg Castle, Loren Brill, Austen Allred, , Matt Doka, Jesse, Jason Levin, David Walker, Shail Patel , Reuben Torenberg, Danika Koenig, Bisla, & others.
David J Phillips126,439 views • 1 year ago
Every voice AI has the same problem The moment you stop talking, it starts talking But humans don't work that way In real conversations, pauses don't always mean interruption. Sometimes they mean: keep going Amit Yadav spent years building real-time AI systems at Meta Reality Labs for Ray-Ban smart glasses Now he's building Fern: a sales copilot that assists teams live during meetings Not after the call ...During it The hard problem isn't the answers anymore. It's timing. Turns out the hardest thing to teach an AI isn't what to say It's when to stay quiet. 🎙️ Amit Yadav (Amit Yadav), CEO & Co-Founder of Fern (ishikilabs), on Fondo.com START Huge thanks to Rho for making this episode possible!
David J Phillips34,155 views • 5 months ago
Global markets were built for everyone. Access wasn't. Millions of traders outside the US watch Tesla move. Watch Nvidia earnings. Watch SpaceX rumors explode timelines. But when it's time to act? The door closes. Mochatrade is rebuilding that layer. A perpetual futures platform for US equities. Built for global traders first. Trade TSLA, NVDA, AAPL, and 50+ US stocks Up to 50x leverage 24/7 markets No US brokerage account required Deposit in local currency. Trade in minutes. Stay self-custodial the entire time. Because the future of trading won't look like legacy brokerage apps. It'll be: Borderless Always-on Crypto-native infrastructure underneath Simple UX on top Global traders still can't easily access leveraged US equities. India alone drives 60% of the world's derivatives volume. That contradiction became the company — starting with India, then the rest of the world. "We have experienced the pain point ourselves" Parth and Chetan grew up in India watching this door stay closed. So they built the platform they wished existed. Not another broker. An access layer for the next generation of retail traders. 🎙️ Parth Maheshwari & Chet Manda, Co-Founders, on Fondo.com START Huge thanks to Comp AI for making this episode possible!
David J Phillips26,703 views • 4 months ago
Most brands don't know if their marketing is really working. They launch campaigns. Spend real money. Watch the dashboards. Then hope. Months later, a report arrives explaining what happened. The campaign is over. The budget is gone. The market has already moved. That's how brand tracking has worked for years. Henk and Harry are repeat founders with exits behind them - two decades building brand trackers for the world's largest brands on one side, a startup built and sold plus a market research career on the other. They'd watched the same problem from the inside for years. Tracking was expensive. Slow. Built around snapshots. Not how brands actually grow. So they built Timelaps: An AI-native brand tracker that continuously surveys thousands of real consumers and shows whether your marketing is moving the metrics that matter. Brand awareness. Consideration. Preference. Category ownership. Brand growth. Research-grade tracking. Updated continuously. Running in days instead of months. At a fifth of the cost of traditional tracking. From "What happened?" to "What's happening right now?" 🎙️ Henk Pretorius & Harry, Co-Founders Timelaps on Fondo.com START Huge thanks to Eragon for making this episode possible!
David J Phillips22,277 views • 4 months ago
A phone call used to mean trust. Now it could be AI. Detectif.ai is building the next layer of cybersecurity for the AI era Not another chatbot Not another AI wrapper Foundational models built to detect audio deepfakes It started with something personal. Tarini's grandfather lost thousands of dollars to a simple deepfake. That moment became the reason to build. Tarini Sai Padmanabhuni had already been in machine learning since grade six Nearly a decade studying how AI works She saw where the technology was headed - and what was missing Fraud has existed for a very long time Cybersecurity has existed for a very long time Deepfake detection is that same discipline extended into the AI world. Today, DetectifAI serves clients across banking, finance, security, telecoms, and OEMs An API that integrates into existing enterprise infrastructure And on mobile, it goes even deeper - into the OEM layer itself. Which makes the vision tangible. You answer a call. Your phone gives you a rolling update on whether that voice is likely AI or a real human Detection during the call. Not after the money is gone. The scale comes from where it's deployed. A single OEM integration can unlock protection for millions of people. And deepfake detection is only the first step. The larger vision is an anti-fraud landscape that helps people navigate scam calls, robocalls, and anything that falls under the purview of fraud. The next generation of AI companies won't just build intelligence They'll build trust 🎙️ Tarini Sai Padmanabhuni, Founder & CEO, Detectif.ai on Fondo.com START Full ep:
David J Phillips15,176 views • 2 months ago
99% of session replays are your app working exactly as expected (hopefully) The other 1% is where the friction hides The bugs. The failed onboarding flows. The conversion leaks. No team has time to watch them all... That's the problem Michael and his co-founder kept hitting at their last company They had analytics. They had session replays. They had ideas to improve conversion. The small fixes always lost to bigger priorities... So they pivoted and built CodeCanary: An AI product engineer that connects to GitHub, PostHog, and Slack It watches your session replays, finds the UX friction, and opens pull requests with the fix Always on. Always shipping. Not dashboards Not alerts Actual fixes 🎙️ Michael Egan, CEO & Co-Founder, CodeCanary on Fondo.com START (full ep in comments) Big thanks to PostHog for making this episode possible!
David J Phillips21,731 views • 4 months ago
As a founder, Emma Ryan Lawler felt like SOC 2 was a major distraction from building. But when she started selling to enterprise logos, it became something that was impossible to ignore. Now, as Product Lead at Rippling, she channeled her first-hand experience of getting SOC 2 compliant to rethink how all automated compliance frameworks are run. It started with a retro. Every other SOC 2 vendor works the same way: Detect a problem Alert your team Fix it...somewhere else The problem with this checklist is that "somewhere else" requires dozens of additional integrations. Because these vendors are just reporting layers on top of your stack. Rippling is your stack. Which means most of your compliance evidence is collected before you start AND when there's a gap in compliance, Rippling fixes it. Unencrypted device --> encrypt it. Wrong app access --> de-provision automatically Incomplete security training --> gate system access until its done. Because compliance really isn't about the report. It's about operating your company securely -- and proving it without a yearly fire drill. Emma shares more about how her own founder journey influenced building Rippling's powerful new automated compliance product, available today for SOC 2 Type 1 and 2. 🎙️ Emma Ryan Lawler, Product Lead, Rippling on Fondo.com START Full ep:
David J Phillips15,649 views • 3 months ago
The biggest hedge funds have entire teams building AI. Most asset managers are on their own. But closing that gap isn't about buying more AI. It's about a problem most people get wrong. You can test code by running it. You can't test an investment by buying it. Software engineering and investing aren't the same problem In software, we've moved from autocomplete to agents to long-running agent swarms - each step climbing another level of abstraction In investing, human judgment still has to exist at the object level So the challenge was never deploying agents The challenge is giving them the context to be useful Every source of knowledge Internal and external Structured and unstructured Connected to the people making decisions. That's what I.M.J. McInnis and his co-founder Ben Finch built at WithAI Not another AI tool A customized environment where agents can work the way investors actually do. Ian knows the problem firsthand. Before co-founding WithAI, he worked at the center of AI and investing at Bridgewater, studying how AI would reshape markets and knowledge work. His conclusion after trying the alternative: "The DIY approach ends in pain" The future isn't simply managing swarms of agents It's knowing what to delegate What to automate And where human judgment remains essential That's why WithAI focuses on amplifying investors rather than replacing them Research faster Monitor more companies Surface information that would otherwise be missed "AI for asset managers, built with you": Security, infrastructure, and white-glove service for harnessing the latest agents 🎙️ Ian M.J. McInnis (I.M.J. McInnis), CEO & Co-Founder, WithAI on Fondo.com START
David J Phillips17,985 views • 3 months ago