<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Kevin Connolly on Medium]]></title>
        <description><![CDATA[Stories by Kevin Connolly on Medium]]></description>
        <link>https://medium.com/@kevinconnolly?source=rss-4e548c9b1d0e------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*QF1NmJ5eUHkPIw7y8xzdiQ.jpeg</url>
            <title>Stories by Kevin Connolly on Medium</title>
            <link>https://medium.com/@kevinconnolly?source=rss-4e548c9b1d0e------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 09:27:43 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@kevinconnolly/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Marc’s Matrix and the Only Two Stories That Sell]]></title>
            <link>https://medium.com/@kevinconnolly/marcs-matrix-and-the-only-two-stories-that-sell-c68deb5d1b95?source=rss-4e548c9b1d0e------2</link>
            <guid isPermaLink="false">https://medium.com/p/c68deb5d1b95</guid>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[cognitive-bias]]></category>
            <dc:creator><![CDATA[Kevin Connolly]]></dc:creator>
            <pubDate>Wed, 12 Jun 2019 13:34:23 GMT</pubDate>
            <atom:updated>2019-06-12T13:34:23.910Z</atom:updated>
            <content:encoded><![CDATA[<h3>Marc’s Matrix and The Only Two Stories That Sell</h3><p>When Marc Andreessen is part of a conversation, I listen closely. As the name behind the Silicon Valley venture capital firm Andreessen Horowitz (a16z) and popularizer of tweet storms, amongst many other accomplishments, he’s got a reputation for sharing worthwhile thoughts. When I heard him on the a16z podcast episode<a href="https://a16z.com/2019/05/07/thinking-in-bets-innovation-uncertainty/"> <em>Innovating in Bets</em></a>, a discussion on the topic of high-stakes decision-making under uncertainty, I took note of his comment that the news fixates on certain narratives. The slippery topic of how those narratives influence almost everyone’s decision-making was touched on throughout the discussion, so I thought I’d try to bring it together here.</p><p>In the mid ‘90s, during the ramp-up to the dot com boom, Andreessen spoke with the veteran reporter Jared Sandberg (then at the Wall Street Journal, now senior executive editor at Bloomberg). It was the early days of the internet, but the new technology was rapidly changing society and the business world. Nevertheless, Andreessen was frustrated. He wanted to know why the internet revolution was not being covered in the news. Sandberg responded with a lesson in journalism: “The story of ‘something is happening’ is not an interesting story. There are only two stories that sell newspapers: one is <em>Oh the glory of it</em>, the other is <em>Oh the shame of it</em>.” The mere fact that internet companies were <em>happening</em> was not fit to print.</p><blockquote>“The story of ‘something is happening’ is not an interesting story. There are only two stories that sell newspapers: one is <em>Oh the glory of it</em>, the other is <em>Oh the shame of it</em>.”</blockquote><p>On the surface, Sandberg’s reality check carries some humor, but a deeper truth in Andreessen’s decades old recollection still applies today: we thrive on stories, particularly the dramatic variety. You don’t have to look too hard to find the Glory or Shame spin because that spin drives sales. But what does this draw to the sensational have to do with making decisions?</p><h3>Matthew Ball (@E3 Jun 10-14) on Twitter</h3><p>Apparently Netflix will die by the end of the year unless a miracle happens per @DisruptionHedge. Literally the most extreme views possible. Miracle or extinction. Gross 172,000 views.</p><p>To get better at something you have to observe mistakes, get feedback and work on improvement. If you want to make better decisions, you probably ask yourself “<em>Was that a good decision?”</em> often. This is where stories come in. Outcomes of events drive most stories, and these stories act as proxies to evaluate success. Thus we tend to judge decisions based on outcomes alone. As a simple example, you can be right for the wrong reasons (i.e. lucky) and, unfortunately, conclude “<em>That was a good decision!” </em>If you are lauded in any way, you’re all the more likely to draw a positive conclusion and assessment of your decision making prowess. Learning to separate judgment of a decision from the outcome, examining “decision quality,” is one of the primary topics in the podcast. To help frame the judgment problem Andreessen offers up what the show’s host referred to as “Marc’s Matrix.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/930/1*uLOV6VcmgUUs58wwLfsCTA.jpeg" /><figcaption>Perception &amp; Judgment: decision quality is often ignored, substituted by the two stories of Genius and Moron.</figcaption></figure><p>For most decisions there is a prior consensus with which you either align or go against, and, in time, your decision will be evaluated as either right or wrong¹. Together, this gives four possible outcomes in the model, but Andreessen’s emphasis is that people tend to collapse the story into <strong>only two narratives that matter</strong>. If you align with consensus opinion, regardless of outcome, there isn’t much to talk about². You’ve effectively joined a<strong> group decision</strong><em>,</em> and the outcome (whether it’s in your group’s favor or not) may generate some news, but it won’t be news about you. However, <strong>when you go against the consensus there is potential for drama</strong>. Now there is an individual to cheer or ridicule, an individual to be revered as a genius, or labeled a complete moron — if you happen to be a CEO or in major league sports, <em>this</em> is when headlines are made.</p><h3>12up on Twitter</h3><p>Did the Raptors just take the single dumbest timeout in NBA history?</p><p>Thus, Andreessen maintains, there are really only <em>two</em> stories in decision making, and they both sell. The problem isn’t only the amplification of these narratives³, but that deeper analysis is rarely present. If you’re right for the wrong reason, you should want to know that. <strong>Furthermore, some outcomes aren’t a binary right or wrong, but rather a measure to some degree.</strong> If you make a decision and anticipate a +30% change for the better, but find a +60% result — you’ll celebrate the success, but you should seek to understand why you were off by a factor of two. Look for flaws in your reasoning and examine the <strong>directional quality</strong>.</p><p>In the absence of support and practice, humans are innately bad at carrying out effective analysis of decisions⁴. As a result, most people cannot manage exploring high-risk, high-reward opportunities. Andreessen suggests one can intellectualize an understanding of these challenges, but the reality of going against the consensus and being wrong is often too much to bear. He wraps up his outlook by likening such an experience to touching a hot stove as a child, such that one learns to never do it again. His conclusion is perhaps more cynically humorous than helpful. After all, Daniel Kahneman’s magnificent <a href="https://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow"><em>Thinking, Fast and Slow</em></a><em>,</em> which has arguably catapulted the whole field of behavioral psychology and cognitive biases into the public mind, was only published in 2011. Kahneman’s 500-page distillation covering decades of research remains part of popular discourse, which suggests there’s still a lot of learning going on.</p><p>Personally, I have a lot of optimism⁵ for improving group decision making, and I don’t think I’m alone. If you’re looking for a place to start, Kahneman’s <a href="https://hbr.org/2011/06/the-big-idea-before-you-make-that-big-decision">12-point checklist</a> makes for a bite-sized introduction.</p><p>[1] An entire category of failure actually stems from decisions <em>not </em>being evaluated. This can be due to timescale (outcomes far from decision), complexity (large, interconnected systems obfuscate a lot), laziness or all of the above!</p><p>[2] Of course, being consensus-wrong or -right <em>should</em> be worthy of reflection and decision evaluation. But the discussion here is about why that often isn’t the case.</p><p>[3] Andreessen comments in the discussion that the two-stories reality is so extreme and so common that it has driven him to mostly withdraw from reading tech news.</p><p>[4] The topic at hand is known as <a href="https://en.wikipedia.org/wiki/Outcome_bias">outcome bias</a> (slightly different from <a href="https://en.wikipedia.org/wiki/Hindsight_bias">hindsight bias</a>), coupled to the amplifying effects of social groups or major news media.</p><p>[5] No escaping biases! See <a href="https://en.wikipedia.org/wiki/Optimism_bias">optimism bias</a>.</p><h4>Additional Recommendations</h4><p>I hope to have the pleasure of introducing these recommendations. Rather than fire off a list of <em>all </em>the books related to decision making, cognitive biases and all that, I tried to pick a short list you might not have come across. Enjoy!</p><ul><li><a href="https://www.oaktreecapital.com/docs/default-source/memos/2014-01-16-getting-lucky.pdf?sfvrsn=2"><strong>Getting Lucky</strong></a><strong> by Howard Marks:</strong> If you’re interested in finance and luck, I highly recommend this memo. It is humorous, concise and insightful.</li><li><a href="https://www.amazon.com/dp/B017I2M8ZC/"><strong>Success and Luck</strong></a><strong> by Robert Frank:</strong> A thought-provoking, slender book on meritocracy, luck, and tax reform. Maybe that’s not the best way to sell it, but it has some nice numerical simulations, cognitive science and societal commentary.</li><li><a href="https://www.amazon.com/Wiser-Getting-Beyond-Groupthink-Smarter-ebook/dp/B00O4CRR9C/"><strong>Wiser</strong></a><strong> by Sunstein &amp; Hastie: </strong>Another beautifully concise book more in the applied than theoretical category, covering the topic of group decision making.</li><li><a href="https://theknowledgeproject.libsyn.com/getting-better-by-being-wrong"><strong>Getting Better by Being Wrong</strong></a><strong>:</strong> Shane Parish of the Knowledge Project interviews Annie Duke, the subject of a16z’s podcast discussed in this post, but in a much longer format. I found this episode more educational and enjoyable.</li></ul><p>Thanks for your time, and if you’ve made it this far perhaps you’d like to read my short piece on <a href="https://medium.com/@kevinconnolly/effort-neglect-and-the-second-textbook-434a3400be2f">learning how to keep on learning</a>. Enjoy!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c68deb5d1b95" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Effort Neglect and The Second Textbook]]></title>
            <link>https://medium.com/@kevinconnolly/effort-neglect-and-the-second-textbook-434a3400be2f?source=rss-4e548c9b1d0e------2</link>
            <guid isPermaLink="false">https://medium.com/p/434a3400be2f</guid>
            <category><![CDATA[education]]></category>
            <category><![CDATA[learning]]></category>
            <dc:creator><![CDATA[Kevin Connolly]]></dc:creator>
            <pubDate>Fri, 17 May 2019 04:55:51 GMT</pubDate>
            <atom:updated>2020-08-08T05:08:18.106Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Second Textbook</h3><p>Sometimes there’s a story you find yourself telling over and over again. Not some sort of <em>thrilling</em> story, but one that’s akin to revealing a shortcut to a fellow traveler on their way to a common destination. I have one such story, on the topic of effort and learning. I call it <em>The Second Textbook.</em></p><p>I studied physics in college and around my junior year courses started getting into esoteric topics which required more effort. These upper level courses typically had additional resources on the syllabus, and often I would find myself in the library reviewing suggested references.</p><p>For one course in particular, solid state physics, I found what I considered my second textbook. The professor had assigned a primary textbook, but at the library I had found a <em>better</em> book. Each week, I’d read a chapter from the primary text, have some wobbly understanding of the topic, and then move on to my second textbook to refine my understanding. The second textbook was so much better that I decided to tell the professor about it, thinking he might appreciate my student perspective and consider using it next time he taught the course.</p><p>He had a friendly laugh when I made my recommendation. It turned out that the textbook I had found was one he had used as the primary text last time he taught the class. At that time, some students insisted they had found another textbook which was much better to learn from. The professor took their feedback to heart and decided to teach from that textbook next time. That book was now the primary text for my class.</p><p>Which book was the better learning resource? My professor had been teaching long enough to have seen the effect of <em>sequence</em> in teaching material. In general, students that were putting in the extra effort tended to prefer the alternative recommended resource. He could swap the order from year to year and it didn’t matter, there were always students that insisted the primary textbook was inferior to the alternative resource.</p><p>Knowledge and skill development happens in stages, accumulates, and comes from various sources and experience. People talk about having different learning styles, usually identifying a preference for text, audio or video. People certainly learn in different ways, but I’d be willing to bet that almost nobody learns in only <em>one</em> way. As for the second textbook, the lesson that always stood out for me is recognizing the <em>ongoing effort</em> of learning.</p><p>Wherever you are in your learning, whatever method or book you’re amidst when you transition from a sense of struggle to a sense of comprehension, you’re likely to attribute that “last click” as the way you learn, or “the best” resource. In this regard, the second textbook can get too much credit, and it can be easy to neglect the fruits of your prior efforts.</p><p>The second textbook, as a concept, as a reality, is <em>always</em> there. Don’t forget to look for it, but don’t forget to appreciate your efforts and give yourself some credit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*P35BuoPkoMBB3-LN" /><figcaption>Photo by <a href="https://unsplash.com/@freestocks?utm_source=medium&amp;utm_medium=referral">freestocks.org</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=434a3400be2f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Top 100 AI Startups]]></title>
            <link>https://medium.com/@kevinconnolly/the-top-100-ai-startups-9136ade072a4?source=rss-4e548c9b1d0e------2</link>
            <guid isPermaLink="false">https://medium.com/p/9136ade072a4</guid>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <dc:creator><![CDATA[Kevin Connolly]]></dc:creator>
            <pubDate>Thu, 14 Feb 2019 08:02:49 GMT</pubDate>
            <atom:updated>2019-02-14T08:26:34.375Z</atom:updated>
            <content:encoded><![CDATA[<h3>Where the money is going &amp; where AI is succeeding</h3><p>Recently, CB Insights published their annual review of what they call the “AI 100,” a list of the most promising AI companies. The ranking is a product of their blackbox assessment over several factors such as patent activity, investor profiles, market potential and competitive landscape, so keep in mind that there are many more companies out there that didn’t make their selection. You can read the ranking details and download the data from their <a href="https://www.cbinsights.com/research/artificial-intelligence-top-startups/">summary page</a>.</p><p>As for this writeup, I wanted to take a look at the broader industry trends, specifically where success was being found with applications of contemporary machine learning, deep learning and, to use loose language, AI¹.</p><p>Given the available data segmentation from CB Insights I had two primary questions and assumptions:</p><ul><li><strong>Where is the money going?</strong></li><li><strong>Where are AI solutions succeeding?</strong></li></ul><p>The first question, as a crude proxy, appeals to me as a measure of venture capitalist knowledge. VCs throw their money where their expertise, analysis and speculation leads them to believe the future of business is headed. You can think of their investments as a partial revelation of their knowledge.</p><p>If you consider that the people building new business, namely entrepreneurs in charge of innovative endeavors and applications, operate from a different set of knowledge and value analysis than VCs, the number of businesses in a given sector can be thought of as another indicator of market value from a second, distinct perspective. In traditional terms, you can think of this competition as a proxy for market validation. If you’ve read Moore’s <em>Crossing the Chasm</em> you might recall the discussion around the importance of competition in driving later-stage adoption of new solutions.</p><blockquote>In sum, the pragmatists are loath to buy until they can compare. Competition, therefore, becomes a fundamental condition of purchase.<br> — Geoffrey Moore, Crossing the Chasm</blockquote><p>With these points in mind, I was curious if the number of businesses that are finding success in different sectors was proportional to investment from the VC side in those sectors.</p><h4><strong>Where is the VC Money Going?</strong></h4><p>Below is a chart of the aggregate VC funding grouped by sector. Remember, this is VC funding into “top” companies applying “AI” in their respective sectors. Clearly, enterprise tech stands out, which isn’t too surprising given how encompassing that sector is. But Government? It’s obvious in hindsight, but for a fun little test why don’t you go ahead and ask yourself where you think the government investments are going (we’ll come back to that shortly).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FA60gChhQbVgzZRy47CEAQ.png" /></figure><h4><strong>Where are the Entrepreneurs going?</strong></h4><p>Next up we have the number of distinct businesses per sector. Given the preamble above about VC knowledge and entrepreneur knowledge, if they were roughly identical you might expect to find a commensurate number of active competitors per sector, regardless of the estimated total addressable market (<a href="https://en.wikipedia.org/wiki/Total_addressable_market">TAM</a>) per sector.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*brsHXAH5ulfmtEMiLzOEzw.png" /></figure><p>If you look at the <em>shape</em> of this chart, you’ll see it doesn’t quite match the shape of the VC investments chart. The dynamics at play here involve not only CB Insight’s selection process, but also the interaction between market size, technology, and funding availability. Let’s explore this a little bit more.</p><h4><strong>Who is winning in what sectors?</strong></h4><p>Considering the VC investment into AI applications in various sectors combined with the number of active competitors in each sector, another interesting question to ask is what is the average funding per org per sector?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S_Gt3fDNumPSAQsRmBi5uQ.png" /></figure><p>Looking at this chart, we see something that possibly suggests a competitive activity to VC knowledge correlation, but there are a few clear outliers.</p><ul><li><strong>Media:</strong> What’s really driving this down is the lack of VC money. There are a few potential interpretations: perhaps AI applications to the media industry are still developing in their early stages (technology as a barrier), or that VC awareness of the potential is lagging, thus investments are low (knowledge distillation as a barrier). Another possibility is that the CB “top” ranking is dubious at best (garbage in, garbage out).</li><li><strong>Industrials:</strong> I don’t have much to say here, but it seems that this sector is getting both a decent chunk of money while possessing only a few leading companies. Most likely, this is a capital intensive sector, or perhaps this is a sector that could use more competition given the willingness of VCs to invest. I’d be curious to know more about the solutions and cost of entry, and if there are any network effects² in this sector that might be feeding into a winner-take-all outcome.</li><li><strong>Government:</strong> In this case, there is both an abundance of funding and few successful competitors, resulting in a very large average funding per business. As hard as it may be to provide solutions for the government, the bounty looks to be well-worth the effort. (You have been pondering what areas these solutions might focus on, right?)</li></ul><h4><strong>How many paths into a given sector?</strong></h4><p>As noted above, sectors like “Enterprise Tech” are quite broad, so much so that they lose meaning. The areas of focus within sectors, provided by the CB Insights data, help reveal these dynamics. For example, the unique number of focus areas listed can be thought of as the number of entry points to a sector (or possibly the number of sub-markets within a sector, depending how you want to view it and how the focus areas relate). Furthermore, this is likely a lower bound: there are <em>at least</em> this many focus areas to compete in, and entrepreneurs are certainly developing more methods to enter any given market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4QgJYRPUN503VdvcvkHylg.png" /></figure><p>If you want to frame your product and business thinking around needs and “jobs to be done,” you can think of this as the primary number of “jobs” per sector that have managed to find traction with AI-based solutions. Taking a look at the government sector in detail reveals that out of the six companies operating in that space, there are only two focus areas: security and disaster management (Alas, revelation! I’m guessing <em>you</em> guessed security, but not disaster management). It turns out that five of the six companies in the government sector are focusing on security. Considering the advancements in computer vision tasks such as facial and object recognition, this isn’t too surprising. But it is interesting to recognize, just to throw an idea out there, that there’s the whole realm of disaster management that might not be getting adequate attention.</p><p>This chart serves to remind us of the obvious: almost all sectors have multiple problems worthy of undivided attention. Despite the temptation to say “company X does AI for Y,” there are often multiple, specific problems to be solved within each of those sectors, and those problems can define entire businesses. Don’t forget to look into the different angles of approach.</p><h4>So, what next?</h4><p>Although there are some caveats involved in examining a subset of companies like this, the overall trends and features serve to motivate a few interesting questions. From the CB Insights data it appears there is success being found in all the major sectors, so the question of interest might not be “where” AI is succeeding, but “how” and to what extent are AI applications succeeding? There appears to be a mad rush in security applications, the enterprise sector appears to be attracting the most entrepreneurs, and other massive sectors like media and telecom are awaiting grand success. These are each worthy of a deep dive.</p><p>The interplay between the availability of funding, the state of incumbents in a given market, and, most excitingly (to me), the role of breakout technologies in the creation of new markets, is fascinating. For the sake of brevity, I didn’t get into the regional data and a few other dimensions of interest, but there is plenty more to be said. If any of these topics interest you, feel free to leave a comment. I’d be delighted to share notes, references or have a casual chat. And if you’ve read this far, thanks for your time in reading my first Medium post!</p><h4>Footnotes</h4><p><strong>[1] </strong>The use of the term “AI” within the research and engineering community is a bit of a hot topic. There’s been some clear mismanagement of expectations, overpromises, and blatant abuses of the term. However, it does have its appeal in being only two letters and conceptually symbolic of “all of those various engineering &amp; research efforts,” so I use it is used here with full respect for the nuances therein.</p><p><strong>[2] </strong>For one of the most accessible introductions network effects and their role in business, I highly recommend Anu Hariharan’s (Andreessen Horowitz) <a href="https://a16z.com/2016/03/07/all-about-network-effects/">presentation</a>. Back when that presentation came out, I read all the references and collected them, should you be interested:</p><h3>Kevin Connolly on Twitter</h3><p>Great slides on network effects via @BenedictEvans news letter this week. Slides and the reference urls in one doc: https://t.co/hgLx8wcZHg</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9136ade072a4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>