Robert Chandler

if you want to learn about how we built the company and what makes it unique, https://lnkd.in/eQS2VRKf is a really really well written article about us by Bill Kerr - better than what I have managed to put together so far!

34

7 Comments

As public software companies got hammered over the last week, David Peterson explores what that means in our latest edition of The Angle. Are we witnessing the end of software? Where did all the growth go? Software isn’t dead, but it’s evolving. And building great companies is no longer about riding the wave of the cloud and building products that are a bit better than the incumbent. Founders who want to build great public companies need ideas that break the mold. And we’re excited to partner with such founders. To read David’s column and explore other exciting tech news, please access and subscribe to our newsletter using the links in the comments. 

10

1 Comment

I’d like to share our most-watched YouTube video to date. 🔥 “Building an MVP for Non-Technical Founders” The video was recorded at a betahaus workshop in Berlin in 2020. As I sat down to prepare for it, I remember thinking that I wanted to create something of immense value - a resource for non-technical founders looking to bring their ideas to life. So, I asked myself, what are the most important things that a non-technical founder needs to know to help them turn their ideas into a functional MVP? What are some tricks they can use to do it fast and efficiently? Here’s a snippet! In this excerpt, I talk about the importance of guiding developers when building an MVP, as well as the most important thing at that stage, which is gaining traction. 💡 I’ve linked the entire video in the comment section below! #mvp #prototype #saas #softwaredevelopment #founders #nontechnicalfounder

14

7 Comments

As we build #GenerativeAI into BHyve, we're realising how little clarity our potential customers have about the costs associated with making the "magic happen". It's also why we get requests from customers saying "Creation should be charged. Search should be free". (Deserves a separate explainer on why Google search is "free" but for another time) For any LLM to give an outcome, there's a ton of work it has to do. I came across this very well-structured report from Ernst & Young which puts a nice little framework to explain the costs of Generative AI. Take a read - https://lnkd.in/dWdvAb_8 How are you looking at brining transparency into cost and value discussions around Generative AI? Would love to hear ideas. #GenerativeAI #AIforGood #CostofAI #ROI #Automation #RAG #BuildingBHyve

14

1 Comment

“Do things that don’t scale” - I’ve heard it many times but had to learn the hard way to apply it. When we launched Omnai a month ago, this was my strategy for getting to 100 beta testers: Record weekly podcast episodes, make 5-10 cutdown clips, and automatically distribute them on various platforms so we always had some signal out there. It was the “let’s throw mud on the wall and see what happens” approach, and it was only 1-to-many communication. Well, I grossly underestimated the time and energy cost of editing, cutting down, downloading various formats, writing captions, and scheduling across nine channels. Doing this for two weeks almost had me pull out my hair of repetitive task boredom. I realized I hadn’t followed the advice I’d heard (and given) so often. I have to do things that won’t scale but will move the needle bit by bit. For now, I’ve rolled back that huge content-production setup I thought was our ticket to traction. Instead, I plan to do something simple that I can manage and that will give me energy in my daily work. My new plan is twofold: 1️⃣ Connect and engage in one-on-one dialogues with internationals in Denmark to better understand their pains and perspectives and to spread the word about the solution we’re presenting to help internationals find and consume local news. 2️⃣ Daily posts on LinkedIn that document our journey and share our ideas, aspirations, struggles, and everything in between. This way, our progress is completely within my power to control—it’s a numbers game. Connect and spark conversation with 100 internationals in Denmark. 50 will reply and teach me something. 20 will resonate with our product and be curious to test our free beta. 10 will love it and become paying customers when we exit our beta. Some percentage of the rest might curiously follow and observe our journey through my daily posts—and if they someday become ready, they’ll know where to go. If this strategy fails to get us to 100 beta testers (our next milestone), it’s either because I’ve not reached out to enough people or because they don’t have the problem we believe they do. If you’re one of the curious people, subscribe to our newsletter here: https://lnkd.in/d5BewEFG Being on there means you won't miss anything important - that’s a promise 🙌

13

4 Comments

Well, we all knew this day was coming…an AI has raised its first funding round. I’m so proud of Boardy! 🚀🤖 When you’re building something that feels like it’s straight out of science fiction, you need people who are willing to step into that future with you — and I’m grateful to have found that support. An incredible group of visionary investors are backing the world's first social AI with $3M in pre-seed funding. 💡💼 The right connection at the right time can change your life. In a world where social media is optimized for eyeballs and engagement, we’re doing something different: using AI to help people make real, meaningful connections. Not just more connections—better ones. 🤝✨ Right now, we're drowning in social media connections that feel empty. We’ve got thousands of connections but somehow feel more isolated than ever. The numbers go up, but meaningful relationships don't. That's why we built Boardy—an AI that does something radically different: it gets people talking. Not through DMs or comments, but through actual conversations. 📞💬 And it's already working. Boardy has introduced: A founder to their technical co-founder 👨‍💻🤝👩‍💻 A startup to their biggest customer yet 💼💸 An entrepreneur to their first investor 💡💰 These aren't just metrics—they're real stories of people finding exactly who they needed to meet. When you call Boardy, he'll search his network to find a great connection for you. This isn’t only based on your profile and professional interests - it’s also someone Boardy thinks you’ll get along with based on your energy and vibe. Have you done your call with Boardy yet? In celebration of this announcement, we’re offering free lifetime access to the first 10,000 people who take a call with Boardy (which we'll probably hit today). 🎉 I’d love for you to experience how AI can actually connect us, not just automate us. If you’re ready to try it out, drop your phone number here, and Boardy will give you a call: Boardy.ai 📲 Let’s see where this goes. Thanks for the support HF0, 8VC, Afore Capital NextView Ventures, FJ Labs, Precursor Ventures, Soma Capital, Golden Ventures, Garage Capital, Northside Ventures, Roar Ventures. Thanks for covering our news Dominic-Madori Davis TechCrunch (link below) #AI #StartupFunding #AIRevolution #SocialAI #RealConnections #BoardyAI #AIforHumans #FutureOfTech #NetworkingReimagined #Innovation

764

289 Comments

🚀 🚀 Exciting news! OpenAI Strawberry (o1) has arrived, ushering in a new era of inference-time scaling in production. As Richard Sutton's Bitter Lesson reminds us, learning and search are the only techniques that scale indefinitely with compute. It's time to shift our focus to the latter. Here's why Strawberry is a game-changer: 1️⃣ Reasoning doesn't require massive models. By factoring out reasoning from knowledge, a small "reasoning core" can leverage tools like browsers and code verifiers, potentially reducing pre-training compute needs. 2️⃣ Compute shifts to serving inference. LLMs as text-based simulators can explore multiple strategies, similar to AlphaGo's Monte Carlo Tree Search (MCTS). 3️⃣ OpenAI's foresight on inference scaling is evident. Recent papers support this:   ➡ Brown et al. showed DeepSeek-Coder's performance jump from 15.9% to 56% with increased sampling.   ➡ Snell et al. demonstrated PaLM 2-S outperforming a 14x larger model using test-time search. 4️⃣ Productionizing o1 presents challenges: determining search termination, defining reward functions, and managing tool integration. 5️⃣ Strawberry creates a data flywheel. Correct answers generate valuable training examples, improving future GPT versions. This breakthrough could revolutionize AI development. Excited to see where this leads! #OpenAI #ArtificialIntelligence #MachineLearning #InferenceScaling #AIInnovation #DeepLearning #TechTrends Courtesy: Jim Fan

27

3 Comments

The infrastructure team at Orb thinks a lot about our production oncall rotation, and how to scale and shape it over time. Here are some core operating principles we've landed on that work well for us so far (of course, any such list will be stage dependent, so take in a high-sodium way): 1/ Try to keep most engineers on-call, avoiding the outcome where a small fraction bear the brunt of instability. This helps build a shared mental model of pain points — calibration on issues is important. 2/ Remove all judgement calls possible at 3am; this requires clear playbooks, links to dashboards as entrypoints for investigation, and alignment on when to escalate (oh -- and train people to bias towards escalation; the natural tendency, especially for newer folks, is to avoid it). 3/ At each weekly handoff, explicitly document what's gotten better and what deserves investment. We also rate each shift (using Asana status updates) between 'on track', 'at risk' to 'off track'. The update should reach the whole engineering team, if not the whole company. 4/ We've gotten a lot of mileage out of a separate #oncall channel with a single thread per investigation. We still try to minimize noise in the alerts/pages channel, but this new venue has become an easier place to rope folks in and ask for help. This channel is also something non-engineers can reasonable track. 5/ As an infra (and in other orgs, SRE, platform, etc.) team, you have to dedicate real roadmap time to investing in oncall improvements in addition to each individual oncall doing so. Some projects just don't get done incrementally. We've done a ton of this — from better instrumentation to observability (whether its metrics or database transactions) to ergonomics (where do alerts go, how do they get deduped, etc.).

47

4 Comments

Code Generation & Customisation are often at odds with one another, but it doesn’t have to be.... 👀 At Speakeasy we’ve taken a multi-level approach to customising SDKs without impacting the level of automation: 1️⃣ OpenAPI-based customisation: Wherever possible, we want the OpenAPI spec to be the single source of truth. That’s why we rely on standard OpenAPI fields to drive customisation and OpenAPI extensions when there isn’t a relevant field available. 2️⃣ Config-based customisation: For look & feel changes to the SDK, we have a small config file. This covers aspect of generation that can’t really be defined within an OpenAPI spec. For example, we maintain a config that allows for your SDK to have a totally flattened interface with parameters passed in-line. 3️⃣ Safe code injection with SDK hooks: This is the cornerstone of our platform’s extensibility and something that is critically important to larger enterprises which inevitably have some custom requirement. SDK Hooks allow our users to add custom code into the SDK in a way that doesn’t require any manual maintenance as new versions of the SDK are published. Our team team has done a phenomenal job of building flexibility into the platform from Day 0. If anyone wants to read more on this topic, check out the getting started guide for hooks that our founding engineer, Tristan Cartledge wrote. See the comments 👇 #openapi #codegen #apis

7

1 Comment

In a world drowning in data, most people tread water. Systems thinkers build submarines. Systems thinking is a powerful lens that reveals the hidden connections shaping our world. It's a crucial tool for designers and founders navigating our hyper-connected, data-driven reality. It's how you turn chaos into clarity, whether you're designing a product that puts a dent in the universe or turning your garage-born idea into a billion-dollar business. The Iceberg Model illustrates this beautifully: → Events (What everyone sees above the waterline) → Patterns (The trends hiding in plain sight) → Structures (The invisible architecture) → Values (The bedrock of it all) Most people only see events—the tip of the iceberg. I've seen countless startups capsize because they were too fixated on surface-level symptoms. The ones that thrive dive deep, challenging assumptions and rewiring systems. Systems thinkers dive deeper. Founders and designers, when you hit a wall, don't simply bang your head against it. Work to understand the system and the interconnections between its various components, all with a mind toward uncovering the hidden levers you can pull and the non-obvious leverage points you can push. Your superpower is connecting dots others don't know exist. You've got the submarine. Now chart a course that reshapes the seascape. Your move, systems thinkers.

54

3 Comments

So no doubt everyone has been talking about the latest llama3.1 large language model that has been released by Meta today. It's supposedly equal or better than ChatGPT 4o or Claude Sonnet 3.5. I've just taken it out for a spin on our GPUs and despite all the energy used to built this model it still can't do simple maths! What to know why? Then like and comment below and that'll keep me motivated to finish my next presentation on this topic!

13

3 Comments

🌟 Archistar Partners with Neighbourlytics for a Game-Changing Integration! 🌟 We're thrilled to announce our partnership with Neighbourlytics, starting with the integration of their Activity Hotspots dataset into the Archistar platform. This layer shows where people spend time across neighborhoods, offering a sophisticated understanding of local dwell patterns nationwide. Here's why this is a game-changer for property developers: 1. Rich Insights: Combine Neighbourlytics' Activity Hotspots data with Archistar's planning database for deeper insights into urban dynamics, walkability, amenities popularity, and more. 2. Targeted Development: Use Neighbourlytics' data to target development projects aligned with local preferences and lifestyles, ensuring your projects resonate with the community. 3. Data-Driven Decisions: Make informed decisions backed by real-time data on activity hotspots, community behavior, and lifestyle trends. 4. Competitive Advantage: Gain a competitive edge with access to a unique lifestyle dataset that complements Archistar's planning tools, helping you identify opportunities and optimize project outcomes. Experience the power of data-driven insights with Neighbourlytics now live in the Archistar platform! Sign up or log in today: https://buff.ly/4dBMpZY Join us in shaping the future of property development and urban planning with data-driven precision! #analytics #neighbourlytics #data #insights #neighbourhoods #development

81

5 Comments

A couple of weeks ago OpenAI rolled-out the Realtime API in public beta, letting developers create snappy, speech-to-speech experiences with the GPT-4o model. The Realtime API delivers two clever technologies: low latency, which is essential for smooth speech-to-speech interactions, and a Voice Activity Detection (VAD) model that knows just when to jump in: farewell to the antiquated request-response model. This means building high-quality speech experiences has never been easier. In healthcare, AI assistants will not only be quicker on the draw but also sound more natural, thanks to native speech-to-speech interaction that cuts out the clunky speech-to-text step. Of course, that doesn't mean we should use them for everything... (I regularly share my thoughts on AI advancements and their transformative potential in healthcare on Nabla's newsletter, Healthcare Hoagie: https://lnkd.in/ebXjVvi2 )

89

9 Comments

The two main reasons RAG and Agents aren't everywhere yet is (1) controllability and (2) accuracy. Good news on that front... Today we took a big step forward in our partnership with LlamaIndex , launching Cleanlab TLM natively in llamaindex so engineers can quickly set up RAG systems with measurably higher accuracy than with traditional LLMs. How does it work? We improve LLM response accuracy using the Cleanlab TLM (Trustworthy Language Model) by computing SOTA trustworthiness scores for every output. The scores are used to search the space of answers for automated answer improvement and enable agent routing and hallucination detection. Where does it work? TLM works as a more accurate LLM itself... or it can be wrapped around any other LLM like GPT, Claude, Llama, or your own proprietary fine-tuned LLM to automatically improve LLM accuracy, adding hallucination detection and trustworthiness scores for every output. Grateful to the hard work by both teams in this growing partnership and proud of our engineers at Cleanlab and friends at Llamaindex for their grit and velocity. Stop building less reliable RAG and Agentic systems. Start using TLM natively in llamaindex. #rag #ai #genai #gpt #hallucinations #agents

102

1 Comment

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

What the heck is Pump.fun and why should anyone care? Pump Fun? Lots of connotations can be derived from those two words, especially put in that order. But in crypto, Pump.fun is a newly launched Solana-based marketplace that allows users to create and distribute their own tokens, with the platform used almost entirely for memecoins. For those who have been around markets for a while, this is the type of activity that happens when hype overtakes sanity. Couple that with the launch of celeb endorsed memecoins, ranging from Caitlyn (Bruce) Jenner, Iggy Azalea to Hulk Hogan and Andrew Tate, and it’s all coming together for rough patch in alt coins (alts have been trending lower since mid-March 2024 - a post on that drubbing to follow). The good news is it looks to have come to an end as evidenced by the amount of active and new users on Pump.fun. New daily users peaked in early June at over 30k per day. It’s now under 2k per day. Add to that total daily users are now sub 8k, a huge drop from nearly 70k a couple weeks ago. See embedded chart (props to @gryphsisacademy on Dune). Further, many memes have been destroyed. Take $JENNER for example. It’s down 94% from its late May peak. Well known memes like Pepe ($PEPE), Bonk ($BONK) and Dog Wif Hat ($WIF) are down ~50% in less than three weeks. As we pointed out a few weeks ago, the first ever memecoin conference was held in Lisbon at the end of May, with it selling out at ticket prices of $850. These type of events and demand for such events are driven by little more than hype. As portfolio managers of tokens we deem to have utility, we welcome this clean out as it drags quality projects down with it and presents buying opportunities for quality tokens. Lastly, many, including us, have said it’s a positive thing that memecoins help onboard new users and in turn bring capital into the crypto ecosystem. We still believe that. But when the scammers, hypesters and D-list celebs takeover, lessons need to be learned. Think the past few weeks have done just that. #crypto #cryptocurrencies #memecoins #memes #digitalassets

5

Xata.io, Inngest, Prisma and Clerk.com launched their summer hackathon on Friday... And I am 100% in. Over the weekend I sank a solid 24 hours in to it already.. safe to say Im enjoying it. So heres what im building ... 🥁 A Chrome extension that allows the user to summarize any Youtube Video. It will then allow those summaries to be viewed, searched and chatted with in a super interactive extension! If youre like me and consume WAY too much youtube.. then you may find yourself forgetting what video it was that mentioned X or Y... especially when its a podcast. The goal of this extension is to help me stay more productive by: ⏰ Letting me know if the content was worth my time with a helpful summary and snapshots 🔎 Allowing me to search and easily reference material from educational poddy c's etc. 📖 Understand which chapter of a video has the best information 🧠 Keep my skills sharp and get ready for when the time comes to build Trigify.io's chrome extension. Super excited to share a short demo with everyone later this week.. so ive successfully managed to consume the videos and generate the summary.. now its all about that sweet sweet UX. Stay tuned for more. To enter yourself visit: https://lnkd.in/eG5GFC_Y

14

3 Comments

🚨big news from Emperia 🚨 Virtual worlds are notoriously hard to build - especially when it comes to running a high-fidelity experience on the web. They require a lot of asset optimisation, compromising on quality. Custom coding for interactions and commercial functionalities. And large teams involved in deploying each experience, which often takes months to create and deliver. Over the last 5 years, we’ve solved all those problems for our own 3D teams. We’ve built the underlying technology that enabled Emperia design, build and publish virtual experiences at scale. We always focused on visual quality and performance, and built the platform with the end user in mind. Today we’re officially opening our tools up for the creative community to build on Emperia platform. Built on UE5, Creator Tools enable: ✨Render and publish cross-device virtual experiences straight from Unreal Engine with no asset optimisation required ✨Plug-and-play UI, enabling any 3D artist to create a fully functional experience with no dev support required ✨ Procedural generation tools and lighting presets to create stunning environments in minutes ✨Commerce modules to support e-comm integration for sales & marketing experiences ✨Customizable interactions from gamification to token-gated rooms with an SDK to expand functionality even further …and so much more! Shoutout to my Emperia team for making this happen, and our partners at Sawhorse Productions, Cominted Labs, Polycount.io, virtuall and many others who helped us refine our creator platform in beta over the past few months 🙏🏻 Simonas Holcmann Emma Preti Charlie Adams Jadir Lino Tomas Tamulenas Tomas Gavelis Zunaib Imtiaz Stephen Cherla Shahzaib Asim Marcelina Slabuszewska #UE5 #EmperiaCreatorTools #builtwithEmperia #3D #unrealengine #AWE2024

214

19 Comments