Siddharth Dhulipalla

Had so much fun recording this episode of Jamstack Radio where we got to jam out about local-first software, SolidJS, and other components of our real-time collaborative stack. Brian Douglas is a long-time mentor and friend and one of the most knowledgeable people about new trends and technologies in the industry. https://lnkd.in/etCYRBhG P.S. if you want to get started with open source or get insights into open source projects, you have to check out Brian's company, https://opensauced.pizza. There's nothing that compares.

16

2 Comments

Yesterday, my team at Vareto announced the first-ever multiplayer modeling platform for Finance. 😍 Now that I can finally speak publicly about what we've been working on, the first thing I want to share is a screenshot from a live multiplayer session we did internally at the company. One of the interesting things about building enterprise software as a startup is that we're still too small to *need* our own product. So instead of doing what our customers do--expense budgeting, headcount planning, etc.--we use Vareto's multiplayer modeling to forecast fun things like total pet spend. Yep, that's right. We had 40 people go into a Vareto model and enter in bottom-up inputs for how much they expect to spend on their cats and dogs this year. 😂 While this isn't very useful--especially since many people entered in clearly exaggerated numbers--sessions like this helped prove that our new 2.0 platform can handle even "extreme" usage situations. (In reality, even if some enterprise customers have thousands of users in Vareto, only a handful are ever likely to be in the same model page at the same exact time given granular data permissions.) You can see more useful, realistic ways that finance and business teams can use Vareto multiplayer here: https://lnkd.in/gV9gqKwz

123

22 Comments

This week Jam hosted 15 DevRel for an off-the-record dinner Here's 7 quick bites about what we learned about The Art of DevRel: 1. The best DevRels work closely with engineering and product. It’s not just about sharing what they're building with devs, it’s also about sharing dev feedback with product. When you get that feedback loop right, that’s how you get magical developer experiences. 2. DevRel should help the entire engineering team do DevRel Like how HR enables any manager to hire, DevRel should enable any engineer to write blogs, create tutorials, give talks, etc. Swyx’s “DevRel as a Service” post was big topic of discussion— will link it in the comments. 3. No 2 companies do devrel the same. No 2 people do devrel the same way There is a lot of art here that’s not science. It’s about finding the right fit person for your exact company Like how there is founder market fit, there is a devrel company fit too 4. Reporting up was a huge point of discussion The best devrels over-communicate internally in the company what is happening on their side Communication creates trust, and trust creates freedom for devrel to do things that don’t scale that builds real community (and results). 5. DevRel is always on; you are always representing the company Community never takes time off, devs around the world are building nights, weekends and holidays You have to love helping people build - truly - to love devrel (and it’s so inspiring to meet people who do) 6. The best DevRel build genuine relationships They become friends with community members and devs The best community and relationships aren’t measurable But can have immeasurable impact on the trajectory of a product/company 7. Last one: Are developers on LinkedIn? Consensus: yes - yes, they are.

20

3 Comments

Today we're launching our newest candidate quality-of-life feature on Ropes - IDE themes. There's a universal dislike for the rigid IDEs that assessment co's force on candidates. These code editors are slow, clunky, and missing the styes/shortcuts that developers enjoy. Developers are unique, and we work in unique ways. Ropes offers two new options to let candidates keep their style: (1) First, many employers on Ropes allow candidates to work in their own IDEs entirely. We've developed a way to gather insights like you were in the room, while allowing developers to use their daily environments. (2) For cases that need a browser-based IDE - we're giving the ability for candidates to choose their style/editor setup before the assessment starts - so they're comfortable when it begins. --- My bet is that if we build a great experience for candidates, we'll be building a great platform for evaluators, too. I hope you'll follow along as we pursue that mission - and we need help! We're hiring founding engineers in NYC - if you're interested, I'd love to talk to you.

47

Apple’s contact-sharing crackdown Apple almost certainly implemented its contact-access changes in iOS 18 to dismantle these growth schemes. Given the use of a prompt mechanism, it’s tempting to view Apple’s contact-sharing crackdown as redolent of Apple’s App Tracking Transparency (ATT) privacy policy. But the comparison ends there. ATT served as a drag on the entire digital economy by constraining growth that would have otherwise been re-invested. Apple’s contact-sharing crackdown will simply abate the short-term windfalls that derive from successful implementations of contact harvesting growth tactics.

16

1 Comment

Thank you for sharing this update on Safe Superintelligence (SSI). It’s incredible to see such significant investment in a startup that’s still in its early stages. The potential for developing safe AI systems is indeed exciting. I agree that leveraging existing computing power, especially in engineering colleges, could be a cost-effective strategy. Utilizing these resources can significantly reduce the overall expenditure and make the most of idle hardware. This approach could be a win-win for both investors and educational institutions. Looking forward to seeing how SSI progresses and contributes to the AI landscape. #SafeSuperintelligence #AIInvestment #MicrosoftAI #GoogleAI #IBMAI #AIResearch #TechInnovation #FutureOfAI #AIStartups #ComputingPower #CostEffectiveAI

3

In late 2021, I became fascinated by the potential of a rapidly emerging technology: generative AI (... even though it wasn't called that yet). I tried to learn about it as much as I could as quickly as I could. As a software builder, I could see just how much generative AI will transform my field. We saw early promise in technologies like GitHub Copilot and projects like AlphaCode, but it was clear that this is barely even the beginning. I was picturing humans and AI collaborating together at building software, humans *augmented* by AI. The idea of software developers augmented by AI is coming closer to reality day by day, even though there is still a long way to go. Today, we've raised a $227M Series B at $977M valuation for Augment, a startup working towards this vision. I am very excited about the team that came together behind this vision (including Guy Gur-Ari, Scott Dietzen, Evan Driscoll, Dion Almaer, and a dream team of Research Scientists, Engineers and "best of" team members in other roles), the investors who've chosen to back us, and most importantly, the product we are building. https://lnkd.in/gEejvzaE

248

54 Comments

At ScreensAI, we're focused on driving as much accuracy as we can with LLMs. But for contract review and playbook execution, it's often the case that vague instructions and subjectivity are as much to blame for undesirable results as the LLM's reasoning ability. Say you want to use an AI system to help determine if the following standard is met across thousands of contracts in your repository: 📜 The contract should limit the vendor's liability to 12 months’ fees. From our recently published Screens Accuracy Evaluation Report: "While any given individual contract reviewer may have a reasonable understanding of what they’d expect to pass or fail this standard, we see many edge cases in practice. If preferences for those edge cases are not clearly articulated, the LLM is left to assume and make a judgment call. Consider these scenarios: - What if there are major exceptions to this liability cap that weaken it so substantially that it might as well not exist? - What if this liability cap doesn’t apply to beta features? Or some specific set of features? What if the service is free? - What if the liability is capped at either 12 months’ fees or some fixed amount, whichever is greater? Or whichever is lesser? What if the fixed amount is very large? Or trivially small? - What if there is a secondary cap or a backup cap that applies under certain conditions at a much different amount than 12 months’ fees? - What if the cap is only for fees paid for the services that gave rise to the liability rather than fees paid in general under the agreement? In any of these cases, the LLM is left to assume. The leading LLMs are capable of reasoning through this type of nuance, but if they are told they must make a decision as to whether or not the standard passes, they might not make the same decision that you would if you don’t expose your preferences." 📊 If you want quantitative aggregate results to assess the risk of a portfolio of contracts, thousands of verbose "it depends" responses won't cut it. Contract professionals have strong opinions, institutional knowledge, and well-defined playbooks. ScreensAI helps them articulate this to the platform up front, systematically validating the types of LLM-powered results they'll get for the given set of expressions of their preferences. Then they can deploy this expression of preferences at scale in perpetuity. 🧙‍♂️ We think this is where a lot of the magic happens in LLM-powered contract review.

21

5 Comments

On Monday, Anthropic released Computer Use API for Claude, along with an upgraded version of 3.5 Sonnet. I spent the last 72 hours tinkering with the Computer Use API. Here's what I learned: 1️⃣ It feels like magic. Watching it run for the first time is genuinely sci-fi worthy. If GPT-4's launch showed us "sparks of AGI," this feels like burning embers. Seeing an AI autonomously use a computer, making real-time decisions and executing actions, makes the future feel tangibly closer. 2️⃣ BUT, it's still early days. The limitations are clear: restricted login capabilities, considerable costs, and reliability issues. Those viral demos you're seeing? Most of them took dozens of attempts to get one clean recording. The technology is impressive, but it's far from production-ready. 3️⃣ Error rates compound in very noticeable ways. Once Claude makes a mistake – like misclicking or landing on the wrong page – it struggles to recover. There's no elegant way to backtrack, and one small error cascades into a series of increasingly confused actions until you're completely off course. 4️⃣ It's going to take a lot of scaffolding to make this valuable for enterprise. The raw capability is awesome, but enterprise adoption needs more: robust privacy controls, human-in-the-loop oversight, domain-specific integrations, and comprehensive security measures. I can definitely see consumer facing applications being dominated by Anthropic and OpenAI, but the greenfield opportunity is in building specialized B2B solutions with focused guardrails and infrastructure. 💡 What we’re building At Mondo, we're focused on exactly this: turning these capabilities into production-ready, enterprise-grade RPA solutions. True self-driving computers are still far away, but the opportunities for targeted, vertical-specific implementations are incredible – and they're here now.

21

2 Comments

Last week, I was all set to head out for my weekly Sunday morning bike ride. Just as I was about to step out, my phone started blowing up with system alerts: - UNUSUALLY HIGH RATE OF REQUEST FAILURES - MEMORY USAGE APPROACHING THRESHOLD - POD FAILURE (x10) "Please let this be a false alarm," I thought optimistically. No such luck. As it turns out, Cozy Earth, one of our premier clients, had launched their $29K ski and spa week giveaway. It was going really well—so well that the surge in traffic from the high comment volume was putting serious strain on our systems. Our amazing engineering team sprang into action with scaling things up, and before long, we were all systems go again. Both our engineers and our infrastructure proved just how resilient they can be. Though I've never felt a stronger connection with https://lnkd.in/g46fzDMU 😅 In the end, the giveaway results left us stunned—in the best possible way: - 61K+ comments in 5 days - 11K+ hyperpersonalized auto-DMs - 3K+ email sign-ups driven - 30% conversion rate What's fascinating is how co-branded giveaways have naturally evolved among our Scrollmark customers. We recently connected Shopkick by Trax and Brandefy App for their own co-branded giveaway. The Cozy Earth and Snowpine Lodge partnership was particularly remarkable—both pages gained around 7K followers each during the giveaway window. But what really matters is that Cozy Earth can track real business impact driven by social. When these leads convert, they know exactly where they came from, no guesswork needed. Huge thanks to Kate Greenwood, Colby McDermott, and Weston Clark at Cozy Earth for trusting us to help turn their viral moment into lasting value. And to our engineering team who showed up and delivered when it mattered most. If you're interested in exploring how to build this kind of momentum for your brand, let's talk. My DMs are open.

43

5 Comments

WE TRIED TO REBUILD GITHUB LIVE IN JUST 4 HOURS — IT TOOK 7 AND WE DIDN'T FINISH (cc Will Hopkins YC S24, Jacob Zwang YC S24, Me and Kevin Gugelmann) Rebuilding Github on Freestyle.sh was an extremely hard challenge because rebuilding Git on Freestyle.sh was an extremely hard challenge. Will and Jacob hooked up an in memory filesystem to Freestyle.sh blob storage for the git backend, we ran git in JS with Isomorphic git, and re-implemented parts of the Git HTTP API from scratch (turns out its not well documented.) Kevin and I took the frontend, rebuilding the Github style for our own site, and designing the models that users would consume based off the underlying Git implementation. We planned to stream for 4 hours, but we were all super into it, so we WENT FOR 7! During the Stream we got: - git clone working - git pull working - git push working But at the end, NONE OF IT DID. We had so much going on that we had broken the core. Thats why we're coming back for round #2. Wanna hear more about Will Hopkins work on Fortress? https://fortress.build/ Wanna learn to build on Freestyle? Check out our new blog on the FART Stack https://lnkd.in/gt-mQvVG Wanna watch the 7 HOUR STREAM? https://lnkd.in/gSAq3H_j Our next stream will be Github PART 2, but after that, what should we build next?

93

8 Comments

“... native solutions built entirely around vectors will provide the ‘speed, memory safety, and scale’ needed as vector data explodes.”  “... we do advanced vector search in the best way possible.” At Epsilla (YC S23), we share the same vision and approach. Thus we have been leading the charge with our unique innovations in a 10x faster, cheaper, and better vector search engine, as the built-in foundation for our no-code RAG as a Service Platform. You can experience the accuracy and relevancy from our advanced technologies via a number of high quality AI-powered smart search and chatbot applications below: 1. Smart Search on 180+ RAG Research Papers (https://lnkd.in/eMPdVae5): This smart search app leverages our RAG platform and vector database technology to help user navigate through the massive 180+ papers with ease and keep up with the latest and greatest advanced research on this important topic. . 2. AI Assistant Chatbot for Knowledge Graphs (https://lnkd.in/eeBTFe_Q): Inspired by insights from Mike Dillinger's posts, this AI assistant chatbot leverages our RAG platform and vector database technology to generate highly relevant, nuanced and contextually aware responses to help user stay ahead of the game on topics like LLM, knowledge graph, and GenAI. 3. AI Assistant Chatbot for Taylor Swift (https://lnkd.in/e_m6dB4P): Powered by our RAG platform and vector database technology, this interactive AI-powered chatbot can answer any questions about Taylor Swift, one of the most impactful artists and successful business women in the world. These high quality AI applications underscore the fact that  Epsilla's vector database technology is not just keeping up with the demands of generative AI, but also setting new standards for what is possible. In addition, our no-code RAG platform allows AI app builders to develop and deploy high quality apps like these within hours. Join us as we continue to innovate and lead in this exciting era of AI-driven applications! (https://epsilla.com) #VectorDatabase #RAG #Chatbots #GenerativeAI #NoCodePlatforms

31

8 Comments

"Context is king!" In a recent insightful discussion with Edo Liberty, Founder and CEO of Pinecone, alongside Derrick Harris and Satish Talluri from Andreessen Horowitz, the importance of how vector databases and #RAG can deliver the most relevant context for AI products. At Perigon, we understand that to extract the maximum value from LLMs, providing the right context is essential. This isn’t just about recognizing the subject matter—which many conversational AI applications can manage—it’s also about ensuring temporal relevance. Without it, even the most sophisticated AI can miss the mark, delivering outdated or irrelevant information. Why does this matter? Because in a world where data is constantly evolving, the accuracy and relevance of the information provided by AI must be on par with what we would find if we did the exploration ourselves for it to be trustworthy. I invite you to experience this firsthand through Perigon’s chat feature on our homepage. Powered by our API, it not only captures subject matter accurately but also ensures the information is timely, tapping into real-time, dynamic data from across the web. Check it out and see the difference real context makes: www.goperigon.com #AI #ConversationalAI

13

1 Comment

~40% of Frec’s customers are engineers working in tech, many of whom expect frequent and large capital gains. They also prefer to self-manage their money, which makes Frec an ideal platform. Meet Evan: an engineering leader who upgraded from investing in ETFs to direct indexing. Direct indexing appealed to Evan because it offers comparable returns to his ETFs while saving him money on advisory and ETF fees. “ETFs are true and tested, and giving folks a different way of investing in the market while saving more money is really appealing.” He’s used many other robo-advisors and other fintech tools, but “Frec seems to maximize tax loss harvesting opportunities better than any other product that I’ve used.” He also likes Frec's transparent fee structure.  “I feel like with a lot of other companies, that’s muddled in with the cost of investments,” he told us, “and you don’t really know how much you’re paying or how much you’re not saving as a result. I think Frec does a really good job of laying all of that out and giving you a true understanding of the cost of these investment products.” Read more about his Frec journey here ⬇

45

2 Comments

For those who remember my previous post about the gov(dot)br hack, check out the HackNotice data for the dropbox domain. Leaked records on the domain are way above average. 25 is really bad and you should expect an incident like the current one if it gets that high. If you're responsible in any way for an organization, you've got to be keeping an eye on this stuff. When ZeroHack tells you a hack is coming, it's because it is. Normalize seeing around corners and preventing incidents, instead of getting caught flat footed doing damage control.

4

Enjoyed reading Ernest Davis's critique today (shared by Gary Marcus) of the recent study by Porter and Machery on preferences by many people for AI-generated poetry over the work of famous poets - Davis - https://lnkd.in/eNTfD8fN Porter and Machery - https://lnkd.in/etqJhbs7 The image is my own contribution to the genre, which displays the same characteristics as the examples discussed by Davis: an arresting original on the left and a banal pastiche on the right. One of P and M's conclusions is that "the simplicity of AI-generated poems may be easier for non-experts to understand, leading them to prefer AI-generated poetry and misinterpret the complexity of human poems as incoherence generated by AI." That sounds a reasonable interpretation, though in some respects they go further than that, in ways that Davis criticises. The analogy with legal work is that, on something complex and subtle, such as a written argument in a dispute 1. there's a huge gulf between what looks plausible to the casual, non-expert reader and something produced with expert insight and judgement - in which less is very often more; 2. of course, there's a lot of mediocre hackery out there in law as in poetry; 3. but even mediocre lawyers can mostly engage somewhat with the facts of the matter, and the needs of their clients - they're not just playing word games, most of the time (yeah, I know, feel free to give your favourite counter-examples...); 4. and a really bad legal hack at least has some prospect, admittedly rather slight, of being held to account in some way for failure to engage with facts and law; 5. an open question is whether AI will start to supplant mediocre legal hackery with something worse or whether it will act as a stimulant to an increase in standards by human lawyers. A bit of both, perhaps.

8

8 Comments

IMO one of the most underrated Guardrails AI features is streaming support. E.g. here's streaming in real time to show how you can validate that the LLM output doesn't contain PII. The difference in latency is wild! Code snippet in comments below:

219

13 Comments

👩‍🚀 Fellow Tech People, we're exploring moving away from our custom API/websocket LLM interfacing at LEAGUES in favor of ELL. 🤔 The potential for streamlined API interactions, simplified prompt management, and reusable functions is really appealing. That said, I’m reaching out to the community: has anyone implemented ELL in production yet? I'm curious if ELL is the way to go for long-term scalability and maintenance reduction when interacting with LLMs. And what about Langchain—why choose one over the other? I'd love to hear your experiences and thoughts! #LEAGUES #ELL #Langchain #AI #TechInnovation

7

5 Comments