Aaron Small

🌿 Agentic AI is Like Gardening - Probably... 🌿 Both require careful planning, continuous iteration, and a deep love for the process. They also involve giving up a sense of control. Playing in the dirt connects us with the dirt. Playing with AI connects us to language. Gardening connects us to the dirt and the roots that can transform those amazing nutrients and minerals into life, organizing itself either for the beauty of it, for our sustenance of both. Gen AI and Agentic Systems in particular, connect us to language and mimic organizational systems and collaboration through that language in a way that creates something new. though you have an incredible amount of control over those early decisions, that is to say, where to plant and to weather to plant, or what to build and weather to do it, how they evolve and behave is somewhat out of your control. I have let myself become frustrated when LLM's don't behave the way I want them to or expect them to. With enough patience and self control, they get much farther than you need, even if its not what you expect. Turns out the nutrients in the soil so to speak, the language we have created for thousands of years, is pretty rich... For now... https://lnkd.in/eCWCBtkm #AgenticAI #Gardening #Innovation #AI #ArtificialIntelligence

5

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

AI is reshaping how enterprise software delivers—and prices—value. Companies like Zendesk, Intercom, and Forethought are shifting to outcome-based pricing, charging only when AI autonomously resolves issues. I've been investing in AI for over a decade. The shift to outcome-based pricing in enterprise software is a seismic change. As I told Jon Victor at The Information, software companies now face a critical decision: Does a company decide to change business models at the expense of losing revenue? It's a necessary risk. With cheap access to LLMs, businesses can build their own enterprise apps. Adapt or die. Why this pricing shift matters: → Vendor success ties directly to customer outcomes → Short-term revenue risks enable long-term market expansion → Industry faces widespread pressure to demonstrate measurable value Pitching AI-powered automation software as a clear cost-reduction tool could ultimately lead to greater revenue for vendors. This is because customers may view AI solutions as fundamentally different from traditional enterprise apps, whose business value is often harder to quantify directly. It's a huge market expansion opportunity. Vendors that nail outcome-based pricing could see massive growth as AI transforms enterprise operations. Those that don't? 😬🫣 How do you see this impacting sectors beyond customer support? Leave a comment. And if you're building in this space, send me a message.

35

3 Comments

PSA: Stop using the original OpenAI CLIP weights. I see a lot of people using CLIP for multi-modal retrieval. What many people may not know is that there are 100’s of CLIP models now since OpenAI’s original release. These newer models offer significantly better performance across many benchmarks. For product search related tasks we recommend the following. + Fastest inference: ViT-B-32 [laion2b_s34b_b79k] + Best balanced: ViT-L-14 [laion2b_s32b_b82k] + Best-all round: xlm-roberta-large-ViT-H-14 [frozen_laion5b_s13b_b90k] You might notice these are not the ones that perform the best on the benchmarks. Why? Because the benchmark only test for very specific tasks which are often quite different from search and recommendations. These are ones we have found to be the best for product related search and retrieval tasks and generally very good first choices if you are unsure where to start (even for non-product related search). Here is a hugging face space with all the data: https://lnkd.in/gxu5UYkt Read more: https://lnkd.in/g5_D4ccs

295

22 Comments

Whether I am on a sales call, at a party, or just speaking with friends, explaining the breadth of Viam’s capabilities in words has been difficult. I have, however, found that once I show the platform at work, eyes light up and ideas about the different devices that could benefit from Viam begin flowing. Here are some of my favorite projects to “show off” when I am giving a demo of Viam. What devices in your life would you hack to do the same? https://lnkd.in/eFn6eD52 #software #robotics #opensource

58

Many assumed that the industry would be well on its way to AGI or superintelligence before AI model builders’ efforts to take advantage of scaling laws might hit a wall. Recent headlines are chipping away at that assumption. One question I've been getting in my inbox: is this the era of diminishing returns for AI? We don't think so. Even if an avalanche of new clusters and synthetic data fail to deliver significantly better model performance, there are good reasons to believe that AI is still going to become orders of magnitude more powerful in the coming years — powering a new generation of startups and ultimately productivity growth and positive benefits for mankind. In fact, with agents and reasoning just beginning to scale and thousands of industry-specific models coming online, it's about to get really exciting. Read our latest blog post on the topic: https://lnkd.in/gFhniNX8

154

8 Comments

How to build like Palantir Technologies 🔮 For 20 years, Palantir has been dubbed a "glorified consultancy" trading at tech multiples. Why? Because they send engineers to their clients to figure out what to build. This may seem counterintuitive for a software company, but what Palantir did was actually quite genius. By embedding engineers to shadow operators across hundreds of industries, they encoded how ideal workflows looked like. And, as a result, created a moat for decades to come. With every new project, they deliver better solutions faster, at higher margins. If you're building an AI agent today, you may not have billions to invest in R&D. But if you go and shadow operators for weeks or months and start building agents to solve their daily problems, you might look back in six months and have a business to scale. Shout out to 8VC for this fantastic piece on Palantir’s success—and to Guillaume De Sá for sharing the article. Please feel free to call Elements Agents a "glorified consultancy". I'd take it as a compliment.

14

1 Comment

Are newer smaller models better then GPT-4? (read till the end) The chart shared by Peter Gostev shows his analysis results, demonstrating that smaller foundational models (6B, 7B, and 8B parameters) are becoming increasingly efficient and effective. Over time, these smaller models have surpassed the original GPT-4 in performance (represented by the green line). This is significant because smaller models require less computational power and resources, making them more cost-effective while still delivering high-quality results. What's the Elo Rating? Elon Rating is a scoring system that ranks models based on their performance in head-to-head evaluations. A higher Elo rating means the model performs better in comparison to others. For example, if Model A has a higher Elo rating than Model B, it means Model A is more likely to outperform Model B in similar tasks. The Elo rating system, while useful for comparing language models (LLMs), has several limitations and potential downsides when applied in this context — such as sensitivity to comparison order, lack of transitivity, and reliance on relative scoring. Therefore the need for complementary evaluation methods (e.g., MMLU scores) for a more comprehensive assessment of model performance (holistic view). For example, a model with a high Elo score might excel at generating responses that are stylistically appealing or persuasive but could lack depth in reasoning or factual accuracy. Human evaluators may unconsciously favor models with more fluent, coherent, or stylistically pleasing outputs, even if those outputs lack substance. This can inflate the scores of models with better "style" rather than better "intelligence." Does it mean the results are irrelevant? Absolutely not! We have learned that for specific situations where factual accuracy isn't critical, we can employ smaller models that require MUCH less computational resources (energy consumption) and obtain stylistic results comparative to LLMs. Note: This post has been written with the help of AI.

4

Why Palantir Technologies alumni are great at founding impactful companies: • They've solved hard problems at scale. • They know how to build products enterprises & governments actually use. • They've seen firsthand how data can transform industries. We’ve posted about ex-Palantir founders before, but here are even more that can call themselves part of the ‘Palantir Mafia’: Alex Arjun Dhillon is building the future of content intelligence at Outtake. Outtake is automating cybersecurity & legal workflows for some of the world's most recognizable organizations by tackling gritty problems in identity, authenticity, and multi-modal AI. Anneli Scopazzi founded Boulevard Recruiting. After scaling teams at Palantir and Figma, she's partnering with both startups and VCs to build out early teams. John Doyle founded Cape, a privacy-first mobile network with secure 5G/4G coverage nationwide. The mobile carrier also blocks hackers and spam. Youssef Rizk and Dimitris Nikolaou were colleagues at Palantir. They went on to found Wondercraft (YC22), an AI-powered audio creation studio for brands and creative teams. Mihir P.  founded Overstand Labs (YC24), a cross-selling platform for enterprise companies. Their AI product has perfect awareness of all your product features and customers, to ensure no revenue is left on the table. Steve Heitkam co-founded Hence Technologies to help companies manage legal risks from geopolitics. Hence helps firms establish efficient legal processes that support smarter decision-making over time. Nick Noone co-founded Peregrine, a platform providing real-time data integration and visualization to help public safety agencies make better decisions. Shaun Springer co-founded Archetype Labs to transform how businesses build custom software. They help companies create tailored software with their AI-driven, no-code tool without the steep cost of custom solutions.

196

5 Comments

We started InspectMind AI (YC W24) to solve one very specific problem which is automating inspection reports. We knew about this problem because my dad is a structural engineer and we saw him doing site visits and writing reports. We realized we can leverage computer vision and LLMs to free up more than half of his time per site visit. And it turns out a lot of companies do site visits/inspections, including very large enterprises in a variety of physical industries where we can save them tens and hundreds of thousands of hours a year with our software, driving significant ROI and margin improvement straight to the bottom line. Today we are solving this one problem for engineering and construction companies. But our $100B #dethroneAutodesk vision is to build the Palantir of physical industries. While today we are automating their reports, tomorrow we will solve 50 of their practical day to day problems with AI and become an invaluable customer first AI second platform for them, specific to their problems and needs and pain points. As we do that we will further and continually develop and strengthen our platform, since if one customer has a problem we can solve for them, chances are many others have the same or similar problems, so the common platform will lift up all boats. We are always looking for missionaries not mercenaries to come join us on this journey towards this big vision. And 100% of my CEO shares are pledged to Christian causes. If you are a hardcore missionary, come join us and help us take this all the way hyper profitably without reporting to bean counters. This time, no bean counting board, no bean counting VC, no censorship HR can get in the way of doing His work, as they are nothing compared to Him.

38

A common and obvious assumption about generative models (incl LLMs) is that they can only be as good as the data they're trained on. They can't "do better" than that. Can they? This interesting new paper "Transcendence: Generative Models Can Outperform The Experts That Train Them" shows that these models can sometime perform better than any of the experts who trained them! While this was done in the context of chess, there is no reason to believe that it might not be the case in other domains as well. Personally, I especially find this useful when I munge two frameworks from different areas to apply to a problem I have. Doing that myself would be challenging, but having the LLM transcend my abilities to do it gives me a great starting point! I like the term "Transcendence" as well - good marketing :)

21

9 Comments

My epic co-founder Tim Cheung did an excellent job (as always) in his latest post for Factory AI. The article is about how manufacturers can leverage the goldmine they’re sitting on—their data. First, he clears some of the fog on today’s hot topics: AI, ML, Deep Learning, and Gen AI. Second, he shares six relevant use cases, each accompanied by a case studies for each. Check out the link to the post in the comments. Give it a read and let us know your thoughts as always. #Manufacturing #FoodAndBeverage

16

3 Comments

OpenAI just announced a $200/monthly access for ChatGPT Pro. https://lnkd.in/gfacR57t It's offering "scaled access" and smarter models. Right now, their API is pretty affordable, without a $200-a-month bill. It'll be disappointing if they apply new scaling restrictions to everything that isn't ChatGPT Pro. For the smarter models, users have become less sensitive to those differences. It's not that important, in my experience, that the model can solve some very difficult coding problems. Doing that better doesn't mean it handles operational use-cases better. When it fails, it's often because it's missing context, like making mistakes with recently released software because that software isn't in its training data. I'll probably pay for this because I have some scaled client use-cases, but I'd be surprised if there are huge sign ups for such a price point.

38

5 Comments

Tembo is leading in the 'Postgres as a Platform' movement. Their most recent blog post(s) below do a pretty good job of explaining why 'Postgres as a Platform' is rapidly becoming an anecdote for database sprawl.

8

1 Comment

🔥 Hot off the press: “What's 🔥 in Enterprise IT/VC - Issue #406” The race to be the last mile AI platform for enterprises - lessons from Palantir's Bootcamp model (BLG is the new PLG), Glean, and others... Full post here: https://lnkd.in/ednmvtCc "Well, this is exactly what Palantir is doing - fixing the last mile problem, aiming to be the platform from which every enterprise builds every single application. What does that tell you? The value is not in the foundational model or LLM, it’s in the workflow and the last mile to make it super easy for enterprises to conceive, build, and deploy GenAI in the enterprise in a rapid, accurate, cost-conscious, privacy preserving, and secure manner. That is the race of the future and every single company that starts with a wedge solving a specific problem will eventually end up in this race. What’s also interesting is how both of these companies go-to-market. This is real enterprise sales - there is no PLG, freemium play at the moment but…all that being said, both are focused on delivering insane Time-to-value or time to the “Aha” or “Wow” moment for its customers. Look at Palantir’s execution from this past week as it continues to leverage its new model - the AIP Bootcamp - BLG for the win!" Read full post here: https://lnkd.in/ednmvtCc

12

3 Comments

📖 Weekend Reading 📖 This week we had profiles on two highly influential figures in tech: Palantir Technologies CEO Alex Karp, and Anduril Industries founder Palmer Luckey. They are in some ways very different individuals, but in others they are radically aligned. 1️⃣ The first piece, on Karp, comes courtesy of Maureen Dowd in The New York Times. Karp appears to have an odd background to be leading a defense-oriented technology company. But what allows the German philosophy doctor to steer Palantir is an extremely principled and unapologetic geopolitical stance: “we have a consistently pro-Western view that the West has a superior way of living and organizing itself.” An interesting observation: “I think we’re in an age when nuclear deterrent is actually less effective because the West is very unlikely to use anything like a nuclear bomb, whereas our adversaries might…Where you have technological parity but moral disparity, the actual disparity is much greater than people think.” Interestingly, while Karp is vocal in his criticisms of some of the Democrats’ policies, he is firmly in the Biden / Harris camp (as a voter and donor) and often butts heads against Palantir’s chairman Peter Thiel when it comes to politics. https://lnkd.in/dMz7BWiV 2️⃣ Next up, Jeremy Stern (Tablet Magazine) profiles Palmer Luckey, another eccentric, highly principled, endlessly fascinating individual. Luckey, who created the Oculus Rift at 19 in a camper trailer, favors anime and collecting decommissioned nuclear missile bases over Karp’s tai chi. But like Karp, he is steadfast in his belief that the West must continue to out-innovate its geopolitical foes (Iran, Russia, China). He also has similar views on rebuilding deterrence, an idea that is core to the company’s mission: “Anduril was founded on the idea the US government had to be reformed so that it could buy deterrence again—and that the only way to deter wars is to be so clearly capable of winning them that no rival power tries to harm core US interests through the use of force.” Anduril is also backed by Peter Thiel, and here, while pursuing what appears like the same objective, Luckey finds himself on the opposite end of the political spectrum as Karp, having famously supported Donald Trump since his first real electoral bid in 2016. I'd recommend this article even just to get a more holistic view of what happened during that time, when Luckey was ousted from Facebook. https://lnkd.in/dpRpnGRF #palantir #ai #llm #defensetech #americandynamism #ukraine #anduril #peterthiel #vr #oculus #fintech #payments #upi

8

2 Comments

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

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

As the AI hype cycle continues, you're also starting to see the "data is the new oil" make its way back into the zeitgeist. It's undeniable that data is valuable, especially proprietary datasets, but how valuable? Up until recently there weren't data marketplaces so it was difficult to gauge how the market would value a dataset. But now with the Snowflake, and other, data marketplaces we can get a sense of what people are willing to pay for data. If you check out the marketplace another interesting aspect is how many datasets are packaged up public data. For example most of the weather data, https://lnkd.in/euhS4GbB would be available for free from NOAA. We've been experimenting with this as well. We've aggregated all the public data from Dutchie eCommerce sites and made it available on Snowflake, https://lnkd.in/eJRTZd5S BDSA sells a full product based on this dataset, https://lnkd.in/ejMqCWuc Broadly, it feels like there's some interesting opportunities here to aggregate, normalize, and sell public data. Have you used any public datasets that are re-packed into a paid product?

2