Jud Valeski

20 hours in and Hoyo's formidable craftsmanship is on full display with ZZZ. ⭐1.0 release pool of 16 characters. Hoyo games have always prioritized quality over quantity, leaning instead on deep systems to enable diverse team comps / fun theorycrafting. Interestingly, Wuthering Waves also launched with the same # of characters in its 1.0 pool. I'm personally a big fan of this approach, and hope to see more games adopt it going forward. ⭐Female to male character ratio of 3:1. HSR currently stands at ~2:1 and Genshin at ~1.7:1. Every faction shown so far has only 1 male char. Surprising given Hoyo's recent titles have focused more on the "mainstream anime" audience instead of catering only to the core otaku audience. Very happy to see Hoyo honoring its roots here. They cooked and core fans are going to be very pleased with all of the fanservice (e.g. cinema artwork, main menu char selection, etc.). ⭐C6 / Cinema artwork is... IYKYK. They clearly tested it with HSR and are fully leaning into it with ZZZ. Congrats to all you Grace C6 havers. They knew exactly what they were doing. ⭐First Hoyo game with dedicated furries / anthropomorphic races. Ellen, Ben, and Von Lycaon are all very compelling characters. Wakayama Shion does such a fantastic job as Ellen's VA. Hoyo absolutely nailed the launch banner here. ⭐Just like other Hoyo games, ZZZ is more "casual" than it is "core." It doesn't demand much of your time as a single player live service title. For me it's the perfect side game that I can see myself spending 1-2 hours per week on. I don't expect everyone to like it (might be too "easy" for core gamers), but it's a winning formula that Hoyo has perfected for its target audience. ⭐One of Hoyo's underrated special talents is making "core" genres widely accessible from a gameplay standpoint across all platforms. ZZZ plays flawlessly on mobile, PC, and console without diluting the experience at all. Everything from the simplified switching system to the flat level design was intentional to make it incredibly accessible even for someone who has never played action games before. This is something really only Hoyo is capable of pulling off with such finesse. ⭐Really digging the comic / manga style storytelling. Glad to see Hoyo continue to innovate on how it delivers its stories and really looking forward to seeing more of this. ⭐Hoyo's investments into its music / animation departments are really paying off. The release trailer for ZZZ is Hoyo's best yet and their character drip marketing is perfected to the point where I'm basically forced to pull on every banner. In many ways, Hoyo is slowly becoming the Apple of the gaming industry, and their IPs are rapidly expanding the audience for gacha and anime on a global scale. I'm honestly humbled. Shut up and take my money... #zenlesszonezero

22

AI memory is fundamental for conversational AI and AI agents. It is critical for building personalized experiences and improving reasoning. Check out Letta’s course to understand how to use AI memory to enhance the customization and quality of your AI app.

15

“... 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

"If you're relying on AI not becoming 100X better, you're about to get steamrolled." -- Jason Warner, ex-CTO of Github and CEO of poolside How we code and build software is being rapidly transformed with AI. Sitting down with Jason the future looks clearer. Over the next 5 years: 🌏 All 7 billion people can code 🏗️ Search bars will spit out software 🔋 AI will generate its way out of data scarcity 🌩️ Foundation models to become cloud companies I unpack the 5 year frontier of software development in Future of Code: AI-led Software Development https://lnkd.in/gze2QSbu #AI #Software #Code

11

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

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

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

We just concluded Embiggen X’s strategy for 2025. For almost two years, we’ve experienced conversations and priorities shift for large companies - more focus on innovating the core the business and less on non-core innovation initiatives. We’ve seen large global companies in the west close down their innovation labs, corporate venturing units, and let go of non-core innovation teams — or at the very least, decreased spending in venturing activities. This is also true for large corporates in Southeast Asia. Simultaneously, we’ve been getting into conversations with CEO’s asking more questions on AI & Digital Business Transformation. No, I’m not talking about building apps. We’re getting questions around: - How do we rewire the “guts” of our operations for us to be able to “scale profitably”? - How do we fix our “spaghetti & siloed” operations? - How do we make sure that we don’t just “buy more” off-the-shelf tech platforms that don’t talk to each other and complicating our already inefficient ops? - How do we increase our NPS score by transforming our customer experience? - Can you help us build a holistic digital business transformation strategy and help us execute too? - How and where can we use AI to become grow better? This is why Q4 of last year in the midst of business challenges, we decided to shift focus from Corporate Venture Building and spin off an AI & Digital Business Transformation company, Embiggen X (https://lnkd.in/geVM4BCF), from the Embiggen Group. Within less than a year, we’ve expanded our operations in Southeast Asia to Europe and 2024 as our highest revenue & profit generating year. We’ve served companies from Insurance, Supply Chain, Telco, Conglomerates, Construction, and Consumer Services to date. We help from Strategy to Execution. I’m beyond grateful to the operating team that made this new chapter of our journey possible. Embiggen X Managing Partner & COO, Xavier (X) Marzan; Partner & Head of Europe, Nicolas Souvlakis (calling in from Switzerland); Junior Partner & CAO Bien Herrera; Partner & CTO Robert Newgen; Chief Growth Advisor Gary de Ocampo, GICD; Advisory Board Member and Incoming Board Director Cliff Eala; Head of Artificial Intelligence Miguel Testa (calling in from Canada); as well as more partners, clients, and supporters. Big goals & challenges up ahead. Excited for whats to come. Onwards & upwards. 🚀

118

5 Comments

💻 Day 16: Deeper Dive into Streams & File Operations 🔄 Today, I delved deeper into C++ streams and learned how to copy data between files efficiently! 📂✨ Key insights: File Streams (ifstream/ofstream): Essential for reading from and writing to files. Copying Data: Used streams to open, read, and transfer data from one file to another, exploring how std::ios flags like ios::in, ios::out, and ios::trunc work together. Error Handling: Managed file errors, ensuring robust file operations by checking stream states. Learning how to manipulate file streams opens up so many possibilities for data management in programs! 📊 Aman Raj Masai IIT-G x Masai #dailylearning #C++ #FileHandling #CodingJourney #100DaysOfCode #LearningEveryDay

1

🚀 There's a lot of publicity around Ilya Sutskever, former chief scientist at OpenAI, launching Safe Superintelligence alongside Daniel Levy and Daniel Gross. This new startup is laser-focused on creating safe superintelligent systems, emphasising safety and groundbreaking engineering. With a singular mission, they aim to revolutionise AI while keeping safety at the forefront. 🌟🔒 It will be interesting to see this develop as more information becomes available, but the securitisation of AI is such a topical conversation and cause of some problems . What are your thoughts on the future of safe superintelligent AI? #AI #Superintelligence #TechInnovation #FutureTech #SafeAI #Innovation #ArtificialIntelligence

10

Just published the first issue of my new Substack - The Vertical AI Report. The future of AI is Vertical, and I'm excited to share more about how this foundational trend is evolving, hopefully from a little different perspective. As an engineer, and a founder, of a Vertical AI company I'm incredibly passionate about the opportunity that Vertical AI is creating for the world. 🔗 Issue #1 is now live, link in the comment section below. #VerticalAI

2

1 Comment

Austin, Head of Technology at Creative NewsHave you ever encountered the challenge of managing empty lists within large data sets and wondered if compiler optimizations could lend a hand? I recently embarked on this exact journey, determined to squeeze every ounce of efficiency out of my code. Initially, I was skeptical about the impact compiler optimizations could have on something as seemingly trivial as handling empty lists. However, I decided to dive into the realm of compiler flags and see what I could discover. By enabling optimizations such as `-O2` or `-O3` in GCC, I found that the compiler was able to streamline my code, eliminating unnecessary checks and enhancing overall performance when dealing with empty lists. One particularly fascinating technique I learned was loop unrolling. This optimization allows the compiler to reduce the overhead of loop control structures, which, even when dealing with empty lists, can significantly cut down processing time in large data sets. If you’re working with large data sets, I encourage you to explore your compiler's optimization settings. Have you considered what hidden efficiencies might be lurking in your code? Tweaking these settings could be the key to unlocking a more performant application. Give it a try and let me know what results you achieve!

1

Elastic goes open source again! What does that mean? After four years under MongoDB’s SSPL license, which failed to get approved as “open source” by the Open Source Initiative (OSI), Elastic goes open source again. This time rather than re-adopting its permissive Apache 2.0 license the company has gone with AGPL, one of the most restrictive open source licenses. This shift is more about semantics than a real change in business strategy. For Elastic, it means they can officially call themselves “open source” again. However, for most users this change means nothing, if you did not use Elastic under the SSPL, you would probably want to avoid using it under the AGPL. #OpenSource #Elastic #SoftwareLicensing #FOSSAware

2

I really like this PMF Framework from Sequoia. https://lnkd.in/gT2nWttx

25

2 Comments

Anthropic's Claude 3.5 Sonnet is really useful. I've been playing around with Claude's coding functionality and artifacts. Claude Artifacts basically allow Claude to generate content based on your prompts. This includes code, designs and even interactive stuff like simple games in a separate window. You literally watch Claude build what you want in super fast speed! I was explaining to an entrepreneur the difference between venture debt and venture capital. It's a bit difficult to illustrate the difference in equity dilution when a founder raise venture debt. I decided to generate a very simple venture debt dilution calculator using Claude. Checkout my artifact below and please give me ideas on features I should add onto my super simplistic creation. 🙂 https://lnkd.in/g3zBi2cd #AI #Claude #Copilot

2

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

Today, while reviewing Layerpath's monthly changelog process, I had an "aha" 💡 moment: we were telling only half the story! 🤔 Like many product-led companies, we've been feature-focused. But our users don't just buy features - they invest in our vision, approach, and community. Here's what I learned from studying companies I admire (Notion, Figma, Canva) and adding our own Layerpath spin: 1. Multi-format storytelling is key. ↳ Fun fact: As a company that helps others create stunning demos and content, we're putting our own tech to work here! 😉) 2. Product updates should reflect company culture. ↳ Shoutout to marketer Swathika who walks the talk - using Layerpath to create our own help docs and resources!) 3. Community insights drive innovation ↳ Your feedback shapes not just what we build, but how we think. 4. User success stories inspire possibilities ↳ Because real impact > feature lists 5. Transparent communication builds trust ↳ Sharing both wins and improvements builds stronger relationships The best updates aren't just about what shipped - they're about the value creation, problem-solving, and community building. Sometimes, the best playbooks need personalization - what works for others should inspire, not constrain, your unique story—taking this learning forward at Layerpath! What's your take on product updates? How do you prefer to learn about new features and improvements? #ProductLedGrowth #StartupLessons #BuildInPublic #ProductUpdates

18

Ilya Sutskever has just launched a new AI venture called Safe Superintelligence (SSI), aimed at developing safe and powerful superintelligence. This comes one month after he left OpenAI. SSI is co-founded by Sutskever, former Y Combinator partner Daniel Gross, and ex-OpenAI engineer Daniel Levy. The sole focus of SSI is to create beneficial superintelligent AI, which Sutskever believes could emerge within a decade. SSI will be a pure research organization, insulated from short-term business pressures. #ai

2

Life and getting momentum is hard. But here's a hack I've written a newsletter for 3 years, every single week on Substack. I've posted almost daily on LinkedIn. It's no joke. I have a small tribe of people who enjoy my thoughts and insights. Only recently a post went crazy (viral as they call it). But it's not far off many I've written before. What did I learn: ~ Timing is everything. Some posts and blogs I wrote over the years were some of my best work, but it didn't resonate with readers. ~ It isn't always logical. The accountant in me knows that my writing doesn't have a clear monetising strategy, but I did it anyway. ~ You only need 1 YES to change your life, and for people to really be inspired and invest in you. Like with StudioNouveau ~ While you're discovering, you will question and have mental breakdowns. It's ok. ~ Writing about God shouldn't be on linkedin they say. My popular post proves this to be untrue. I'm not ashamed of the Gospel. Anywhere. ~ SAICA and possibly FNB South Africa will potentially partner with me at my OWN speaking event to help other people with ambition and dreams. ~ Find the balance between pursuing your interests and generating cash flow to meet your respoinsibililties ~ Seek to make a positive impact on others. Not status. You don't know how your life will turn out. But consistency, unconventionalism and resilience are life hacks. A big thank you to Chadley Langley CA(SA) Beverley Moore (Cert.Dir®) MBA Laszlo Nolan CA(SA) Thabo Masuku Darryl Roux David Muamba, Professional Accountant (SA) for always reading and encouraging. https://lnkd.in/dXRP2rGA

41

5 Comments