Mike Lyons

Notion vs Glean reminds me a lot of Tesla vs Waymo... Just like I think Tesla is going to solve the 'software' problem of autonomous vehicles before Waywmo solves the 'hardware' problem... I think that Notion is going to solve 'enterprise knowledge search' before Glean solves the problem of document and knowledge creation... Why? A couple reasons: 1) Just like Tesla, Notion has a huge advantage when it comes to the amount of underlying data and documents to train on... They already have 100M+ users, and likely 10x that in terms of the number of documents, files, etc. in team's workspaces. 2) It's very hard to change people's behaviour re: where/how they create 'work'. It's why a lot of companies have come and gone in that category... As exciting as Glean's AI writing assistants might seem, most people are just fine (or even prefer) a good old fashioned editor/writer. I could be way off here, but this will be an intersting one to follow. Thoughts? Where am I getting it wrong? #ai #enterprise #knowledge #management #notion #glean

3

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

NVIDIA's Jensen said "ChatGPT democratized computing, Llama2 democratized generative AI" ...and now Llama3 is the next level 🔥 It's fascinating to watch Meta's strategic moves. With Llama3, they prioritized training on a massive 15T token dataset to pack all into a lean 70B param model, instead of building a massive model. This allows Llama 3 to match trillion+ parameter models like GPT-4, but at 1/10th the compute, storage and inference costs! 💰 This technique was published by Google DeepMind a few years ago https://lnkd.in/gx7VU8aA Meta is not a dark horse in AI anymore. They might be the top dog. https://lnkd.in/gRRJYuxt #llama2 #llama3 #llm #chatgpt #gpt4 #ai #ml

16

5 Comments

Bessemer Venture Partners series on VerticalAI is insanely good. If you haven't read it yet, you should. Link in first comment below 🔽 #VerticalAI

7

3 Comments

When you quantize down an LLM model, can you maintain its accuracy? I used to think no, but the data says otherwise. For example, Llama 3.1 405B FP16 has an average Accuracy score of 86.63. If you quantize down this model from FP16 to INT4, its Accuracy score falls to 86.47. That’s a recovery ratio of 99.81%. 🔥 🔥 In return, you get 4x more VRAM and higher performance. As people struggle to fit their models onto 8x H100s, a tiny loss in accuracy for 4x more VRAM sounds like a great bargain. Secret sauce seems to be the GPTQ algorithm, combined with neural networks being robust to quantization. Source: Neural Magic's Llama-3.1 70B and 405B INT4 models on Hugging Face. The link's in the comments. #opensource #llama3

35

5 Comments

I like this idea of "technical taste". It gets into where software engineering is as much creative art as it is science. Four great takeaways (IMHO) for software devs in this era in particular: 1. Aspiring engineers should cultivate a sense of curiosity, experiment with different tools and technologies, and embrace a mindset of continuous learning. 2. AI has the potential to streamline processes and enhance productivity for engineers, but it may also lead to disruptions in traditional software development workflows. 3. Developing technical taste and judgment is essential for making informed decisions about which technologies and approaches to pursue. 4. Collaboration and open-mindedness are key to leveraging the full potential of AI and staying ahead of technological advancements. I also really liked these two quotes from Sam Schillace: 1. "The right time to do something is when you have that feeling in the pit of your stomach that's like, 'oh, this is a great idea and it's going to suck to build because nothing's ready yet.'" 2. "Technical taste is like, 'how well have you consolidated that set of experiences and heuristics into judgment that you can apply accurately when you see new things?'" 3. "It may be the case that very small teams can do very large projects, or like we were talking about before, it may be the case we're just going to get really ambitious about what we try to do with the same size of teams, which is kind of where I would put my money." Great job on continuing to put out great episodes, Brett Berson and team First Round Capital! #ai #engineering #softwaredevelopment

5

If you are building with LLMs, you can get really far in terms of building a highly accurate and reliable product by optimizing your development cycles for fast iteration: ➡️ Pick an off the shelf model - Llama /OpenAI/Anthropic. ➡️ Religiously track the prompt-model generation pairs in production and patch each edge and corner case you identify through a combination of prompt iteration and retrieval optimization(for RAGs). Avoid getting stuck trying to research the current best model or wondering whether to finetune on your custom dataset. Focused iteration will get you to your ideal outcome faster than everything else.

20

1 Comment

🚨🚨 🚨 Unlocking Innovation: The Power of the Adjacent Possible Innovation programs can stall when the team tries to ideate way beyond their collective perspective and frame of reference. Like a software engineer brainstorming genetic modification of mycelia, or a senior executive ideating process improvements which they haven't been participated in for decades. It's challenging, exhausting, and ultimately self-defeating. Enter the concept of the "Adjacent Possible" - a game-changer for innovation. What is it? The Adjacent Possible, coined by Stuart Kauffman, represents the realm of new ideas within reach based on our current knowledge, resources, and capabilities. Why it matters: ➡️ Realistic Innovation: it focuses on achievable advancements, not far-fetched concepts. ➡️ Incremental Progress: each innovation opens doors to new possibilities. How to leverage it: • Constantly explore the boundaries of your adjacent possible. What's in your line of sight is what you know best. • Combine existing ideas in novel ways. Even small perturbations from status quo can generate large innovation leaps. • Stay curious about developments in adjacent industries. Cross-domain patterns create fresh sources of ideas, while also expanding the adjacent possible. Most of all, it takes the 'genius' pressure off the team. No one knows everything, but each one of us knows a lot about something, and we can all drive plenty of innovation potential by focusing on the spaces we know best. Are you exploring your Adjacent Possible? Share your thoughts below! 👇 #Innovation #innovationsystems #AdjacentPossible

8

2 Comments

💥 Finnt is now publicly available! 🌎 Building an AI-native software is a whole new beast - actually that’s why there’s so many great demos but so few amazing experiences in production. The nature of Generative AI makes it difficult to reproduce identical, deterministic outcomes. Where a traditional feature would have two statuses “functional” vs. “bugged”; in the GenAI world, a slightly different prompt or setup can drastically change the output. But this is also what makes AI so powerful and why we are so excited to be building in this space. 🧠 As a reminder, Finnt specializes in extracting crucial data from large, unstructured, and complex documents — including OMs, PPMs, 10-Q/10-K, as well as prospectuses, reports, leases, loans, etc. It then integrates these findings into AI-generated and templated memos, enriching them with real-time market data. 🔑 Key facts: - Ingest any large, complex, and unstructured files - Design custom frameworks tailored towards your specific use cases - Select your preferred LLMs, from closed to open-source options - Customize the goals, structures, and formats of your output reports - Augment your reports with third-party sources and live market data - Ask anything to your files and transform responses into insights - Seamlessly integrate Finnt into your existing workflows 👉 Check the comments to get the link to signup. PS: you might need to get a custom onboarding but we are nice people 😅

179

47 Comments

Interesting conversation between Jason Calacanis and Tekedra N. Mawakana Mawakana, CEO of Waymo (Google's self-driving project).... (Paraphrased) Jason Calacanis: Why are you not testing your Waymo self-driving on highways? You seem focused primarily on city environments like San Francisco, yet in my experience being a tesla-owner from the very beginning (and using both auto-pilot and then full self driving) it is better on the highways. You seem to be taking the opposite approach. Is it because speed kills? Tekedra: Our mission is safety and to get to Level 4 driving (the highest level of autonomous driving, no driver behind the wheel, no drivers' license) and so we’ve made a deliberate decision to focus on crowded urban environments because that is the hardest problem. Anything that doesn’t contribute to our getting to Level 4 is not interesting to us (not intended, but could be back-handed compliment to Musk 🙂) You don’t learn anything new self-driving on the same boring monotonous highways. --- Great point! #waymo #allin #selfdriving #google

4

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

$2 H100s: How the GPU Bubble Burst Last year, H100s were $8/hr if you could get them. Today, there's 7 different resale markets selling them under $2. What happened? The market has flipped from shortage ($8/hr) to oversupplied ($2/hr), because of reserved compute resales, open model finetuning, and decline in new foundation model co’s. Rent instead. https://lnkd.in/gpQEidmC

2

1 Comment

The newest issue of The Pragmatic Engineer found a surprise uptick in eng hiring in the typically slow summer season - I'm curious to see how the macro environment impacts this uptick as we head into the fall, but I remain hopeful. I asked Gemini to give me the tldr of the article to share with my team (sorry, Gergely Orosz) and found it missed one of the most interesting pieces of the article. Gemini's highlights - Unexpected surge: Tech recruitment has seen an unusual spike in June and July, contrary to typical trends. Increased demand: Startups, especially AI-focused ones, are driving the hiring uptick. Global phenomenon: This trend isn't just limited to the US; Europe is also experiencing increased recruitment activity. Reasons for hiring: Companies are backfilling positions, investing in growth due to improved financials, and responding to increased VC funding. Challenges remain: While the job market is improving, finding qualified candidates and offering competitive compensation remains difficult for many companies. What I found most interesting - The nature of the backfills: the article finds that a lot of new headcount are actually backfills of folks that were "over-slashed" during rounds of layoffs. Not entirely surprising with the "trim the fat" and "rebuild lean" mentality, but interesting nonetheless. Either way, the uptick is encouraging (and don't always rely on Gemini's tldrs). #hiring #startups #VCbacked #softwareengineering

11

1 Comment

YC's "request for startups" - Winter 2025 - Government + Public Safety 1. Build LLMs to automate specific government tasks like form filling, application reviews, and document summaries 2. Create computer vision systems for license plate detection and crime prevention 3. Build software that reduces police paperwork time from hours to minutes 4. Create tools for emergency response coordination and dispatch 5. Build platforms that help communities and law enforcement communicate effectively - Manufacturing 1. Build ML-powered robotics systems that reduce labor costs in US manufacturing 2. Create automation tools specifically for American factories competing with overseas production 3. Build specialized industrial robots for factory inspection and maintenance 4. Develop systems that help manufacturers operate efficiently in US industrial hubs - Chips + Engineering 1. Build LLM tools specifically for FPGA design and optimization 2. Create Al systems for ASIC design that reduce development costs 3. Build tools that optimize specialized computation like crypto mining or data compression 4. Develop Al-powered CAD/CAM software that makes engineering tools more accessible - Stablecoins 1. Build platforms for businesses to hold and manage stablecoins 2. Create tools that make it easy for developers to integrate stablecoin payments 3. Build systems for banks to issue their own stablecoins 4. Develop infrastructure for cross-border stablecoin payments and remittances - New Jobs 1. Create tools that help people run local service businesses 2. Build platforms enabling people to earn income providing services online 3. Create systems that help service providers work more efficiently with Al assistance 4. Develop tools that help small businesses operate at the same level as large corporations #YC

4

There's now no shortage of options for open source model hosting providers. i.e., take your pick for Llama 3 (Groq, Together AI, Telnyx, etc.). Am I wrong in thinking the only real difference now between them is speed, cost and readability? If so, this space just became commoditized quickly and reminds me of the telco provider wars where its a race to the bottom on price. Which means the cost of running AI applications is approaching something close to zero... 🤯 #ai #cx

14

2 Comments

Some investors are still dismissing most AI startups as "GPT wrappers" -- i.e. they think companies that are using OpenAI or Anthropic will quickly get replaced or have no moat, since every company has access to these models. This is a flawed mental model. You can easily see the flaw if you think of OpenAI and Anthropic as providing intelligence, just as humans provide intelligence. You would never dismiss "a company that hires people" as "a human wrapper" and claim there's no moat because anyone can hire people. Instead, you have to think more deeply about accumulating advantages, network effects, product, expansion opportunities, etc. There are almost no YC startups that are literally just building a thin wrapper on OpenAI or Anthropic. They are all building real products, understanding customer problems, developing features that anchor them deeper into workflows, and solving problems that can't just be solved by using ChatGPT. A better general way of thinking about AI startups is The Black Hole Model, which I've talked about here before.

141

28 Comments

Seeing lots of great founders disrupting EDA with AI Most EDA tools don’t do real “engineering,” but rather take the rote tasks out of implementing engineering output from centuries of “experience” Can AI collapse the heuristics of the best engineers that ever lived to actually automate engineering? Hard to beat EDA at its own game by making it better with AI Generating the boundary conditions that engineers normally use to start the engineering process and better optimized at the end would be a great complement to existing EDA

108

22 Comments

As LLMs drive the ease and cost of programming down to 0, we are going to see a lot more software. Tons of niche micro-SaaS companies. But also much bigger platforms filled with features PMs could only dream of in the past. It’ll be a competitive environment. Deep tech will be rewarded more than ever. Lots of value capture will occur at the infrastructure layers. How does anyone selling software compete in this future? Sell picks and shovels.

5

1 Comment

Recently, there's been a lot of talk about Whisper, an open source speech-to-text engine developed by OpenAI. Here is how we use Whisper at Nabla: 1️⃣ The well-documented Whisper issues the article highlights are exactly why we spent 3 years and $5m, gathered and manually annotate a unique dataset of 7000 hours of medical encounters audio, and leveraged that dataset to train our own model. 2️⃣ Our model is based on Whisper, but it contains many improvements specifically developed to suppress hallucinations and make the transcription of medical terms more accurate than any off-the-shelf speech-to-text engine on the market. 3️⃣ In addition, the transcript is not directly included in the patient record. A second layer of processing extracts relevant data from the transcript. Each note produced is split into atomic facts, which are checked via an LLM query against the transcript and the patient context. Only facts for which we find definitive proof are considered valid. 4️⃣ After data is extracted, it is manually reviewed, edited and corrected if needed by the same clinician who did the encounter. Nabla also enforces that the review happens shortly after the encounter. 5️⃣ Each new version of our model undergoes rigorous evaluation by human professionals to ensure the documentation is both comprehensive and aligned with industry standards. 6️⃣ We have processed 9 million medical encounters in the past 24 months and got direct feedback from nearly 10,000 physicians about ways to improve Nabla. While some transcription errors were sometimes reported, hallucination has never been reported as a significant issue.

675

70 Comments