Mark Koltnuk

Unity and Weights & Biases were my focuses for this week's R&D day @ LangLabs #unity3d #weightsandbiases #gamedev #promptengineering #ai AI videos generated by Runway Gen-3 Alpha

10

🆕 💥 BIG UPDATE — getting new backers, releasing AIstats 2.0, and hiring for many positions Me and Eugene Sertukov have been hands down building for many months and are excited to share a sneak peek into where AIstats is going. We’ve analyzed 12M match scenarios and developed the most complete ecosystem of football metrics used by regular football fans. In ~1 week, we will release AIstats 2.0, our vision of the ideal modern livescore app. It's designed for speed and clarity, providing fast and clear match insights with predictive power. Users will be able to quickly scan and understand matches and their potential outcomes, all powered by our comprehensive historical data analysis. We are already supported by a bunch of cool folks: FAANG top managers (including Mariusz Gąsiewski from Google — one of the top mobile experts in CEE), product experts (including Alexander Pavlov from BDC Consulting) and repeat founders (including Alexander Nevedovsky from FundraisingHackers.com). They are all excited about our mission to democratize the average football fan (and later other sports) experience with AI. Also, we are grateful to Veronika Lindorenko for arranging the Ū Hub | Startup Community in Poland events where I met met Eugene and joined AIstats. We’re hiring (remotely!) across the board, so please recommend cool folks we should chat with: 1️⃣ Head of UA, with at least 3 years of experience in traffic acquisition for subscription-based apps, particularly with Meta and Google Ads, and managing budgets over $100,000 monthly. 2️⃣ Senior UI/UX Designer, with at least 3 years of experience in mobile app design, with familiarity in the livescore and betting sectors being a notable advantage. 3️⃣ ASO Manager, with at least 2 years of experience and a strong understanding of subscription app markets. Passion for football will be a great advantage. 4️⃣ Business/System Analyst, with at least 2 years of experience in structuring functional requirements of data-oriented software products. A good understanding of the football and betting industry will be an advantage. 5️⃣ Fractional CTO, an experienced individual who understands how to manage high-load systems, build effective data infrastructure, and has experience in assembling tech teams. We offer competitive comp and equity — with a flexible split. To apply, connect with Alexey Vorobey on LinkedIn or message me on Telegram (@avarab) with the link to the resume/profile and mention why you’re excited about the AI x football intersection. 🙏 Max repost, please 🙏

117

15 תגובות

From 1.5K to 6K in 5 mns time on a 3090. A work(flow) in progress. #ai #flux #comfyui #opensource

77

12 תגובות

🔍 tonari.no: Cutting-edge communication or just an expensive toy? Tonari portals offer life-sized, immersive experiences with 10x the fidelity of video calls, sub-100ms latency, true eye contact, and room-scale immersion. But at $25k or $1k+/month, is it worth it? I was at a startup 10 years ago where we wanted to use TVs and cameras to "merge" two offices so I suppose a market exists. Tonari takes that to the next level, but the cost is insane! Would you choose this over Zoom? Or does this feel like the most expensive creepy TV in your office? 🤔

12

1 תגובה

I did a Quick Test of CMS by Tejas Kulkarni in AR I tested the model available in the 3D showcase assets of the CMS. The result was good, but it needs proper remeshing and retexturing. Unfortunately, these features are only available in the paid plan, and I am currently on a free plan. What intrigues me is the intense competition to commercialize AI in the 3D space. Startups like Meshy, Spline, Rodin, Luma, and many others are already in the market, with more expected in the coming years. There are plenty of open-source repositories like Instant Mesh, Triposr, Unique3D, CRM, and Wonder3D. However, these resources are not user-friendly for everyone to learn and install locally, which is why platforms like Pinokio and others are building Gradio UIs for one-click installs of open-source AI. Comfyi is another excellent tool that offers most of the 3D model nodes needed. However, setting it up can be challenging. The use cases for 3D assets are vast, spanning games, Unreal Engine, Blender, VR, AR, and more. AI has already solved mesh generation, and tools like Stable Diffusion and ControlNets handle retexturing. What remains are rigging, animation, and fixing UV maps. In the AI startup space, many companies quickly commercialize open-source models without fully understanding the value they provide. I'm sharing my thoughts on this: Imagine a 3D modeler who doesn't spend time on social media and is focused on creating amazing 3D models. This could take 10-20 hours or more, depending on the task, experience, and skill set. If they knew they could accomplish this with a single text prompt, how would they feel? They'd have two options: either utilize the technology or be left behind, as companies will prefer those who can complete tasks in minutes rather than hours. If they can dedicate 10-20 focused hours to building a 3D model, or even more, they can definitely spend that time researching and installing open-source 3D models locally without paying any subscription. Many jobs have already been lost. At 24, AI is the most significant evolution I've witnessed since the internet's growth. Companies that excel in marketing, content creation, and running successful campaigns will thrive in the short term until a competitor offers better quality services at lower rates. #ai #ar #xr

5

2 תגובות

As part of our talk with Anh Chu, Solution Architect at Microsoft, we discussed the maturity levels of copilots that we see SaaS companies building to infuse their products and offerings with Generative AI. 💡 Key Takeaways: 🔹 Starting with Fundamental Functionalities 🔹 Progressing to Integrated Assistance 🔹 Reaching More Autonomy

22

4 תגובות

In thinking about what to post here about Pod Engine's new developments, I realized there's a lot I've taken for granted and not mentioned! Something I find particularly cool is the sheer scale of our architecture and ML pipelines. Pod Engine now has tens of thousands of _fully tracked_ podcasts. For each podcast, this means: 1. At least the last 6 months of episodes transcribed and analyzed (for many we've processed every episode in their back-catalog too!) 2. Every transcript fully searchable semantically and by keyword 3. Their own podcast analytics page (attached, will improve a lot this week too!) We actually track millions of podcasts (specifically every active podcast, and a fair few more), and the number of fully tracked podcasts is steadily growing. Essentially our machines have "listened to" many, many hundreds of thousands of episodes and extracted insights extracted from every single one. Super powerful stuff when harnessed appropriately! I'll be adding some extra data to the podcast pages this week and would love some feedback. They'll have: 1. Podcast contact details 2. Overall reviews (Itunes + Spotify) 3. Detailed (individual) reviews for Itunes 4. Various proxies for listenership data (Castbox subscribers etc.) 5. All social links You can access all of this for free today just by logging in to a free account. Our pilot customers will soon have access to lots more though, which I'm extra excited about!

17

We keep rolling out continuous 𝗠𝗼𝘃𝗲𝗿𝘀𝗲 𝗖𝗮𝗽𝘁𝘂𝗿𝗲 𝗦𝘁𝘂𝗱𝗶𝗼 improvements 👣 👋. For this one we've also taken a step towards a much wanted feature, fingers capture 🤘! #markerless #ai #motioncapture #3danimation

29

1 תגובה

🚨 New Video Alert: Flux vs SD 3.5 — Who Wins the AI Battle? 🚨 We just tested Stable Diffusion 3.5 (including both Turbo & Large models) across multiple platforms like HuggingFace, Mage.Space, Civitai, Tensor.art, SeaArt.ai, Glif.app, and GoEnhance.ai. Wondering if this is the Flux Killer? 🤔 We compared outputs from Stable Diffusion 3.5 and Flux models, and the results might surprise you! Watch the video and see which model offers the best image generation in terms of details and prompt accuracy. ➡️ Watch Now: https://lnkd.in/gCy6Afyc 🌟 Also, don't miss our NFC products for digital business cards and more! Visit: Tapit.Club for the latest in NFC tech! #AI #StableDiffusion #FluxModel #StableDiffusion3_5 #MachineLearning #NFC #AIArt #DigitalBusinessCard #ImageGeneration

1

A common concern I hear about generative AI is the potential for creating realistic AI fakes that could be used maliciously. Let me address this and explain why this is not the case at Colossyan. 🔒 The first company I founded specialized in deepfake detection. My team and I are highly aware of the risks associated with this technology and how to effectively mitigate them. 🔒 We have an automatic content moderation system in place to ensure all generated content is appropriate. 🔒 We collaborate closely with our clients to understand their use case and ensure our services are used responsibly, rather than for misinformation or impersonation. We recognize the cybersecurity challenges posed by AI video. However, the solution is not to halt the technology's progress. It’s about having certified experts (like my team) actively managing these risks. And that’s exactly what we do.

29

1 תגובה

I just built an Agreement Redliner Agent in <5 mins with Lyzr AI Agent API Studio. Here's what happened 👇 One of our partners got in touch, mentioning they had a client who needed a agreement redliner agent. While I could share a somewhat similar case study, the partner wanted something even closer to their requirements. That's when I thought, why not try building one? With Lyzr Agent API Studio (https://agent.lyzr.ai/) , I quickly designed an agent with the 'no-code wizard' that can, 1. Check the agreement for any objectionable clauses 2. Redline the clause 3. Share the reason for redlining I gave the agent five key clauses to look at. (described in the app) And voila. I just built an agent, and also an app for the agent, in under 5 minutes. And no code. And not even OpenAI o1. We used GPT-4o. Lyzr's agents have in-built ReAct framework that helps in superior reasoning and execution. And Lyzr agents automatically enhances agent prompts by applying 'chain of thought' structure. How I built the agent - https://lnkd.in/eEpYXMVi Try the agent - https://lnkd.in/eKJY5Ykn Build your own agent - https://agent.lyzr.ai/ If there's a process you're thinking of automating, know that Lyzr lets you do so without chasing your IT team. Got a complex or critical business process or job role to automate, speak to us. https://lnkd.in/eh6ih-9q

87

4 תגובות

Photo + Audio file = Lifelike...

6

Drag-and-drop a subject into an image with a vastly different style achieving a style-harmonized and realistic insertion of the subject 🎉 👉 Magic Insert: Style-Aware Drag-and-Drop Nataniel Ruiz, Yuanzhen Li, Neal Wadhwa, Yael Pritch, Michael Rubinstein, David E. Jacobs, Shlomi Fruchter | Google Abstract We present Magic Insert, a method for dragging-and-dropping subjects from a user-provided image into a target image of a different style in a physically plausible manner while matching the style of the target image. This work formalizes the problem of style-aware drag-and-drop and presents a method for tackling it by addressing two sub-problems: style-aware personalization and realistic object insertion in stylized images. For style-aware personalization, our method first fine-tunes a pretrained text-to-image diffusion model using LoRA and learned text tokens on the subject image, and then infuses it with a CLIP representation of the target style. For object insertion, we use Bootstrapped Domain Adaption to adapt a domain-specific photorealistic object insertion model to the domain of diverse artistic styles. Overall, the method significantly outperforms traditional approaches such as inpainting. Finally, we present a dataset, SubjectPlop, to facilitate evaluation and future progress in this area. 👉 https://lnkd.in/d6QB7d8y #machinelearning

3

This is how the evolution of technology took place expressed in an artistic way! #evolution #arts #generativeAI #Humanevolution #businessdevelopment

1

With Depix Design Lab you can grab a screenshot from any CAD app and bring it alive in seconds using the Style Drive feature. You can give Style Drive a try on Discord. Here's the invite. 🚀🧨 https://lnkd.in/dfzCXV-u > the prompt type /style then hit 'enter' then select your base image, and then select your style image then 'enter'. #Depix.ai #CADVisualization #Engineering #futurerender #easy2use #fun

31

This is the story of the legendary Wordware (YC S24) launch. 𝟏𝟎 𝐝𝐚𝐲𝐬: 𝟕𝐌𝐌 𝐩𝐞𝐨𝐩𝐥𝐞 𝐫𝐮𝐧𝐧𝐢𝐧𝐠 𝐨𝐮𝐫 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭𝐬; 𝟐𝟕𝟖𝐤 𝐮𝐬𝐞𝐫𝐬; $𝟏𝟎𝟎𝐤 𝐞𝐯𝐞𝐫𝐲 𝟕 𝐝𝐚𝐲𝐬; 𝐛𝐫𝐞𝐚𝐤𝐢𝐧𝐠 𝐏𝐫𝐨𝐝𝐮𝐜𝐭𝐇𝐮𝐧𝐭 𝐬𝐞𝐫𝐯𝐞𝐫𝐬 (𝟕𝟎𝟎𝟎 𝐮𝐩𝐯𝐨𝐭𝐞𝐬). This is the whole story of how an 𝐀𝐈 𝐀𝐠𝐞𝐧𝐭 𝐜𝐚𝐧 𝐦𝐚𝐤𝐞 𝐲𝐨𝐮 𝟒𝐤 𝐞𝐯𝐞𝐫𝐲 𝐡𝐨𝐮𝐫 and a guide on how to replicate the launch: Just to remind you: Wordware is the first AI Operating System (aiOS) that enables anyone to build, iterate, and deploy AI with Natural Language. In simpler non-VC trillion-dollar TAM terms, we make everyone an AI Developer. You can run your AI on our front end or get an API and hook it up in your product. With this, you get to great Agents 20x faster cause everyone can iterate, yielding better quality. Like, how the hell can an engineer know what a good contract should look like? And yup, today they are the one who has to iterate on it (or deal with hundreds of requests from PMs/CEOs and rebuild the application again) Try it or sign up for a workshop if you are building in AI: wordware.ai Okay, so back to the story: I have been in startups for over six years, and going viral has happened to me for the first time. Is it PMF? Probably not yet, but it gives us a chance! 𝟏𝟎𝟏 𝐨𝐧 𝐡𝐨𝐰 𝐭𝐨 𝐝𝐨 𝐚 𝐥𝐞𝐠𝐞𝐧𝐝𝐚𝐫𝐲 𝐥𝐚𝐮𝐧𝐜𝐡 𝑻𝒉𝒆 𝑺𝒆𝒕-𝑼𝒑 This might apply only to AI Infra companies. Also, we were primarily lucky, but maybe somebody can plan this out with what we learned: (0.) get yourself some amazing angels - Azeem Azhar, Ben Tossell, Ben Parr, Matt Schlicht, Yohei Nakajima, Shawn swyx W - their advice and help with virality have been invaluable 1. Use your underlying product to launch something viral — in our case, it was twitter.wordware.ai (Kudos to Kamil Ruczynski, our head of growth, for the idea) — an AI Agent that reads all your tweets, roasts and analyses your personality. BTW read my roast here: https://lnkd.in/dinS4Ti9 or our compatibility report with Robert Chandler: https://lnkd.in/dkyzrTTs 2. Plan the launch of your main product (wordware.ai) at the peak of virality. Make it evident that the viral product was built with your tool and redirect interested people to your website. Check out the above links to see how many references to Wordware there are. 3. We made it extremely easy to play with the agent. If you click Duplicate, you land in a fully functioning prompt that you can edit quickly. Try it. This clever technique meant that we redirected a LOT of traffic to wordware.ai. From there, there was an offer to book a free workshop and a link to our ProductHunt Launch. 4. Ridiculous exposure means that even a couple of per cent matters (X was perfect for this as there are more techie people there). 𝑻𝒉𝒆 𝒑𝒓𝒐𝒄𝒆𝒔𝒔 rest of the story+business insights here https://lnkd.in/dxwkqsF2 (I run out of characters) Also we hiring for all positions -- but i must have known you before, dm me on WhatsApp

529

45 תגובות

"'The best part about Autoblocks? It's boring. And that's exactly what you want.' - Zahid Mahmood (CTO at Anterior (formerly Co:Helm)) Hear why making AI testing feel familiar and straightforward is a game-changer for the dev teams building LLM products. Innovation isn't about complexity - it's about making the complex feel simple. That's why companies like Anterior trust Autoblocks AI 🚀

11

Topic 2 Midjourney - Image generation behind-the-scenes Here's what is happening behind the scenes when Midjourney or Meta AI generates a beautiful picture of a young girl eating pizza. It's "diffusion models" doing this magic. Draw a parallel to how you would learn to cook a pizza in real-life. You would do a reverse engineering or "uncook" the pizza into the exact amount of flour, sauce, cheese, vegetables that goes into making the pizza. and then you would follow a recipe to re-cook it. Similarly, diffusion models would have seen billions of images of a pizza and a girl separately during their training phase. ▶️Take pizza in this example: These models would "un-cook" or breakdown these images of "pizza" and to understand all the "ingredients" that went into the existing image. Ingredients in this case would be the distinct attributes of an image like edges, textures, colors, shapes. This process is done a million times to create a foundation model that can generate any kind of image of a pizza. ▶️The model learns how multiple objects appear together. For example, it might learn from images of "people" eating various "foods". From this learned representations, the model is able to place the pizza in a position relative to the girl.   Essentially, there's destruction and recreation to create a foundation model. Post that, it can create any variations - like a boy eating a margherita pizza, or a boy eating a pizza sitting in the corner of a park. #Midjourney #DallE #GenAI #imagegeneration #LLMs

23

2 תגובות

Update from Iteraite - we've added the R.I.C.E Prioritization template to our "Product Manager Toolkit" - as well as made our "Entrepreneur Toolkit" completely FREE 🎉 With the "PM Toolkit", you can go through the entire Discovery of a new product/feature using #AI as both an assistant and feedback generator. It costs less than a cup of coffee a month, and we're always adding more features. ☕️📅 If you just want to test it for free - just use Access Code: freePRD2024 to get access to the PRD (Product Requirements Document) generator, and more. We're not a "chatbot" - so you need to actively work through your Product, with the assistance of AI 🤓 The RICE Prioritization is a good example of that - AI is a great tool to help you think through your prioritization, but in the end, you (the PM) were hired to make the final call. Also - we've made the "Entrepreneur's Toolkit" completely #free! Why? Because we want people to be making better products and startups, and there is no excuse anymore to not knowing how to make an Elevator Pitch or Pitch Deck (with our AI-assisted tools). It's on our dime, so "you're welcome" 😃 Remember - we won't use/sell/see anything you write on our site - it's completely safe... because that's how we roll. 🎳 Just go to our site at iteraite.com to try them out! #freetools #aiforgood #productmanagement #productmanager #strategy #startup #entrepreneur #pitchdeck

15

4 תגובות