<?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 Mihai Raulea on Medium]]></title>
        <description><![CDATA[Stories by Mihai Raulea on Medium]]></description>
        <link>https://medium.com/@mihairaulea?source=rss-4a51aaa081cb------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*cfwCd7O3hkPoFCzN_6H9tA.jpeg</url>
            <title>Stories by Mihai Raulea on Medium</title>
            <link>https://medium.com/@mihairaulea?source=rss-4a51aaa081cb------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 00:32:42 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@mihairaulea/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![CDATA[How to Cryptographically Protect Unpublished Mathematical Work]]></title>
            <link>https://medium.com/openspur/how-to-cryptographically-protect-unpublished-mathematical-work-45d2e4d500ff?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/45d2e4d500ff</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[math]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Wed, 09 Sep 2026 18:41:53 GMT</pubDate>
            <atom:updated>2026-09-09T18:41:53.448Z</atom:updated>
            <content:encoded><![CDATA[<p>I asked. Prompted a little. ChatGPT delivered.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5Hmb-BhTY-ffJCptrvXD7Q.jpeg" /></figure><h3>How Mathematicians Can Cryptographically Prove When Their Work Existed</h3><p>Mathematicians have always cared about priority.</p><p>Traditionally, if you wanted to establish that you had a result first, you relied on things like:</p><ul><li>a dated notebook,</li><li>an email to a colleague,</li><li>a seminar talk,</li><li>a journal submission,</li><li>or, most commonly today, an arXiv preprint.</li></ul><p>These mechanisms work reasonably well once you are ready to disclose your work.</p><p>But what about the period <strong>before publication</strong>?</p><p>Suppose you have been working privately for six months on a difficult problem. You have:</p><ul><li>LaTeX files,</li><li>Lean formalizations,</li><li>code,</li><li>intermediate lemmas,</li><li>failed approaches,</li><li>notes,</li><li>and perhaps the core of an important result.</li></ul><p>You do not want to publish any of it yet.</p><p>But you would like to be able to prove later that:</p><ol><li>this exact body of work existed at a particular time;</li><li>the timestamp did not come from your own computer clock;</li><li>particular collaborators made particular contributions;</li><li>the server storing the work could not read it;</li><li>and the entire history could later be independently verified.</li></ol><p>This is possible today.</p><p>The basic idea is simple:</p><pre>private mathematical work<br>        ↓<br>       Git<br>        ↓<br>signed commits<br>        ↓<br>SHA-256 commitment<br>        ↓<br>trusted timestamp<br>        ↓<br>encrypted remote storage</pre><p>The mathematician still works normally:</p><pre>git add .<br>git commit -m &quot;Prove compactness lemma&quot;<br>git push</pre><p>The cryptography can happen automatically in the background.</p><h3>The core problem</h3><p>There are really four separate questions:</p><h3>What existed?</h3><p>You need a cryptographic commitment to the exact state of the project.</p><h3>When did it exist?</h3><p>You need an independent timestamp that does not rely on your laptop clock.</p><h3>Who created it?</h3><p>You need a digital signature tied to a contributor.</p><h3>Who could read it?</h3><p>You need end-to-end encryption.</p><p>Git helps enormously with the first and third questions.</p><p>But Git alone does not solve the timestamp problem.</p><p>A Git commit may say:</p><pre>September 9, 2026 at 20:41</pre><p>but that timestamp is not strong evidence. The author controls it and can backdate it.</p><p>If you want to prove that a repository state existed at a real-world time, you need an external authority or public system to observe a cryptographic commitment to it.</p><h3>Step 1: Put the work in Git</h3><p>Git is useful for mathematics even if you are not primarily a programmer.</p><p>A research repository might look like this:</p><pre>paper/<br>    paper.tex</pre><pre>lean/<br>    Main.lean<br>    Lemma7.lean</pre><pre>notes/<br>    approach-a.md<br>    counterexample.md</pre><pre>computations/<br>    experiment.py</pre><p>After:</p><pre>git add .<br>git commit -m &quot;Complete proof of Lemma 7&quot;</pre><p>Git records an exact snapshot of the committed state.</p><p>Over time, you might accumulate a history like:</p><pre>Aug 11   first approach<br>Aug 13   counterexample discovered<br>Aug 17   new construction<br>Aug 20   main theorem formulated<br>Aug 22   Lean proof succeeds<br>Aug 24   generalization</pre><p>This matters because a research history can be more persuasive than a final PDF.</p><p>It shows the evolution of the work.</p><h3>Step 2: Do not rely on a normal private GitHub repository</h3><p>A private GitHub repository is private in the access-control sense.</p><p>It is not private in the cryptographic sense.</p><p>GitHub controls the infrastructure and can technically access repository contents under certain circumstances.</p><p>If your requirement is:</p><blockquote><em>The hosting provider should be cryptographically unable to read the mathematics.</em></blockquote><p>then you need client-side encryption.</p><p>One interesting current solution is <strong>FOKS</strong>, an open-source system for end-to-end encrypted Git repositories and team collaboration.</p><p>The important property is that repository data is encrypted on the researcher’s machine before it reaches the server.</p><p>So the server stores ciphertext, not your mathematics.</p><p>Conceptually:</p><pre>Alice ─┐<br>Bob   ─┼── encrypted research repository<br>Carol ─┘<br>             ↓<br>          server<br>             ↓<br>         ciphertext</pre><p>The server can store and synchronize the repository without knowing what is inside it.</p><h3>Step 3: Add collaborators normally</h3><p>Suppose Alice, Bob and Carol are working together.</p><p>You create a team and give each researcher access.</p><p>The important thing is that each collaborator has their own cryptographic identity.</p><p>Do not share one laboratory-wide key.</p><p>You want your history to distinguish:</p><pre>Alice → commit A<br>Bob   → commit B<br>Alice → commit C<br>Carol → commit D</pre><p>If Carol later joins the project, she can be granted access.</p><p>If she leaves, access to future states can be revoked and keys rotated.</p><p>There is one unavoidable limitation:</p><blockquote><em>You cannot make someone forget information they were previously authorized to read.</em></blockquote><p>Removing Carol today can stop her reading future work. It cannot erase a copy she legitimately downloaded last month.</p><p>That is true of every collaboration system.</p><h3>Step 4: Sign every commit</h3><p>Encryption answers:</p><blockquote><em>Who is allowed to read the research?</em></blockquote><p>It does not answer:</p><blockquote><em>Who created this specific state?</em></blockquote><p>For that, sign commits.</p><p>Git supports cryptographically signed commits using SSH, GPG or S/MIME.</p><p>For example, using SSH signing:</p><pre>ssh-keygen -t ed25519</pre><p>Then:</p><pre>git config --global gpg.format ssh<br>git config --global user.signingkey ~/.ssh/id_ed25519.pub<br>git config --global commit.gpgsign true</pre><p>Now:</p><pre>git commit -m &quot;Complete main estimate&quot;</pre><p>creates a signed commit automatically.</p><p>Each contributor should have their own key.</p><p>Later, you can show that a particular repository state was signed by Alice rather than merely appearing in a shared project.</p><h3>Step 5: Create a SHA-256 commitment to the repository state</h3><p>Git already hashes its internal objects.</p><p>For long-term evidentiary purposes, however, I would create a separate SHA-256 commitment over the actual committed state.</p><p>You want something conceptually like this:</p><pre>paper.tex      ─ SHA256 ┐<br>Main.lean      ─ SHA256 ├── repository state hash<br>notes.md       ─ SHA256 │<br>experiment.py  ─ SHA256 ┘</pre><p>The result is a small fixed-length value such as:</p><pre>bd75519bbc5c3a7f...</pre><p>That value commits to the entire project state.</p><p>If one character changes in one file, the commitment changes.</p><p>The crucial property is this:</p><blockquote><em>You can reveal the hash without revealing the mathematics.</em></blockquote><p>A SHA-256 digest tells the outside world almost nothing useful about the underlying proof.</p><h3>Step 6: Ask an independent timestamp authority to timestamp that hash</h3><p>This is where the system becomes much stronger than a normal Git history.</p><p>Use an <strong>RFC 3161 Time-Stamp Authority</strong>, usually called a TSA.</p><p>An RFC 3161 TSA does not need your files.</p><p>It receives a cryptographic digest.</p><p>Conceptually:</p><pre>private repository<br>      ↓<br>SHA-256 commitment<br>      ↓<br>RFC 3161 TSA<br>      ↓<br>signed timestamp token</pre><p>The TSA sees something like:</p><pre>bd75519bbc5c3a7f...</pre><p>It does not see:</p><ul><li>your theorem,</li><li>your proof,</li><li>your Lean source,</li><li>your notes,</li><li>your collaborators’ comments.</li></ul><p>The timestamp authority digitally signs evidence saying that this digest existed no later than a particular time.</p><h3>Step 7: Use UTC as the authoritative time</h3><p>If you care about proving not only the date but the exact hour and minute, RFC 3161 is particularly useful.</p><p>A timestamp token includes an authoritative UTC time.</p><p>For example:</p><pre>2026-09-09 17:49:31 UTC</pre><p>You can display it locally as:</p><pre>2026-09-09 20:49:31<br>Europe/Bucharest<br>UTC+03:00</pre><p>But the cryptographic source of truth should remain UTC.</p><p>A good proof record might show:</p><pre>Trusted timestamp:<br>2026-09-09 17:49:31 UTC</pre><pre>Local display:<br>2026-09-09 20:49:31 Europe/Bucharest</pre><pre>Repository state:<br>bd75519bbc5c3a7f...</pre><pre>Contributor:<br>Alice Smith</pre><pre>Signature:<br>VALID</pre><p>That is much stronger than a locally generated date.</p><h3>Step 8: Automate everything</h3><p>The system will fail in practice if researchers have to remember a complicated security ritual every time they discover a lemma.</p><p>The ideal workflow is still:</p><pre>git add .<br>git commit -m &quot;Main estimate works&quot;<br>git push</pre><p>Behind the scenes, the system should automatically:</p><pre>1. read the new commit<br>2. calculate the repository SHA-256<br>3. sign the state<br>4. request an RFC 3161 timestamp<br>5. verify the response<br>6. store the proof<br>7. encrypt the repository<br>8. push it</pre><p>The researcher should not have to think about ASN.1, timestamp tokens, certificate chains or Merkle trees.</p><p>Cryptography should be infrastructure.</p><h3>Step 9: Timestamp every commit automatically</h3><p>Git supports hooks that run after certain operations.</p><p>A post-commit hook can start the timestamping process automatically.</p><p>Conceptually:</p><pre>git commit<br>    ↓<br>commit created<br>    ↓<br>SHA-256 state calculated<br>    ↓<br>RFC 3161 request sent<br>    ↓<br>timestamp proof stored</pre><p>The proof should live outside the working tree, for example:</p><pre>.git/attestations/</pre><p>Do not commit the timestamp file into the same repository state you just timestamped.</p><p>Otherwise you create a circular problem:</p><pre>commit<br>  ↓<br>timestamp<br>  ↓<br>commit timestamp<br>  ↓<br>new commit<br>  ↓<br>new timestamp</pre><p>Keep the evidence outside the committed tree.</p><h3>Step 10: Make git push the hard guarantee</h3><p>There is one fundamental limitation.</p><p>Suppose Alice creates a commit at 14:03 while her laptop is offline.</p><p>She reconnects at 16:17.</p><p>No external cryptographic system can honestly prove that the outside world observed her result at 14:03.</p><p>Her laptop clock is not independent evidence.</p><p>The strongest externally defensible statement is:</p><blockquote><em>The result existed by 16:17.</em></blockquote><p>For this reason, I would use the following rule.</p><h3>On commit</h3><p>Try to obtain a timestamp automatically.</p><h3>On push</h3><p>Require a valid external timestamp.</p><p>So:</p><pre>git push<br>    ↓<br>does every pushed commit have a timestamp?<br>    ↓<br>yes → continue<br>no  → timestamp first</pre><p>That creates a very useful invariant:</p><blockquote><em>Every state that ever reached the shared repository had already received an independently verifiable timestamp.</em></blockquote><h3>Step 11: What happens if there is a priority dispute?</h3><p>Suppose five years later someone disputes who had a result first.</p><p>You reveal:</p><pre>Git history<br>+<br>repository states<br>+<br>contributor public keys<br>+<br>RFC 3161 timestamp tokens</pre><p>The history might look like:</p><pre>Aug 11  Alice  proposes construction</pre><pre>Aug 14  Bob<br>finds obstruction</pre><pre>Aug 18  Alice<br>modifies construction</pre><pre>Aug 21  Carol<br>formalizes main lemma in Lean</pre><pre>Aug 22  Alice<br>main theorem verifies</pre><pre>Aug 26  Bob<br>simplifies central estimate</pre><p>For the August 22 state, an independent examiner can reconstruct the repository state and calculate:</p><pre>SHA256(repository state)<br>=<br>bd75519bbc5c3a7f...</pre><p>They then compare that with the timestamped commitment.</p><p>If it matches, the conclusion becomes:</p><blockquote><em>This exact repository state existed no later than the time contained in the independently signed timestamp.</em></blockquote><p>That is much stronger than:</p><blockquote><em>Alice says she had the proof on August 22.</em></blockquote><h3>What this system proves</h3><p>It can give strong evidence of several things.</p><h3>Existence</h3><p>These exact files existed by time T.</p><h3>Integrity</h3><p>The files shown today are the same files that produced the previously timestamped commitment.</p><h3>Chronology</h3><p>Different research states can be ordered through independent timestamps.</p><h3>Attribution</h3><p>Contributor signatures show who signed particular states or changes.</p><h3>Confidentiality</h3><p>End-to-end encryption prevents the repository host from seeing the plaintext research.</p><h3>What it does not prove</h3><p>Cryptography cannot solve every priority dispute.</p><p>It does not prove:</p><h3>Mathematical correctness</h3><p>A timestamped false proof is still false.</p><h3>Originality</h3><p>Someone could have copied an idea five minutes before timestamping it.</p><h3>Independent discovery</h3><p>Cryptography cannot determine where an idea came from.</p><h3>Legal ownership</h3><p>Ownership can depend on employment contracts, funding arrangements and jurisdiction.</p><p>What the system gives you is something narrower and still extremely valuable:</p><blockquote><em>verifiable contemporaneous evidence of possession and development.</em></blockquote><h3>For important work, add a second anchor</h3><p>RFC 3161 relies on a trusted timestamp authority.</p><p>For exceptionally important research, you can add another independent mechanism such as <strong>OpenTimestamps</strong>, which anchors cryptographic commitments into Bitcoin.</p><p>The architecture becomes:</p><pre>┌── RFC 3161 TSA<br>                      │<br>repository state ─────┤<br>                      │<br>                      └── OpenTimestamps<br>                              ↓<br>                           Bitcoin</pre><p>The two mechanisms provide different kinds of evidence.</p><p>RFC 3161 gives you an explicit trusted time.</p><p>Bitcoin gives you a decentralized historical anchor.</p><p>They complement each other.</p><h3>You do not need one timestamp per file</h3><p>Suppose your research repository contains 5,000 files.</p><p>You do not need 5,000 timestamps.</p><p>The entire state can be reduced to one cryptographic commitment:</p><pre>5,000 files<br>    ↓<br>SHA-256 state commitment<br>    ↓<br>32 bytes</pre><p>Timestamp those 32 bytes.</p><p>The same idea scales further.</p><p>An institution could combine thousands of research commitments into a Merkle tree and timestamp only the root:</p><pre>project A ─┐<br>project B ─┤<br>project C ─┼── Merkle tree ── root<br>project D ─┘</pre><p>Each project receives a small inclusion proof.</p><p>Millions of private research states could therefore be externally timestamped using only a small number of actual TSA requests.</p><h3>Why this matters especially now</h3><p>For most mathematics, an arXiv submission remains an excellent mechanism for establishing public priority.</p><p>But arXiv requires disclosure.</p><p>There may be a long period before that:</p><pre>first idea</pre><pre>   ↓</pre><pre>private development</pre><pre>   ↓</pre><pre>important breakthrough</pre><pre>   ↓</pre><pre>formal verification</pre><pre>   ↓</pre><pre>collaborator review</pre><pre>   ↓</pre><pre>paper preparation</pre><pre>   ↓</pre><pre>arXiv</pre><p>That private interval can last months or years.</p><p>Historically, mathematicians have relied largely on notebooks, emails, colleagues and institutional trust during that period.</p><p>Cryptographic timestamping adds another possibility:</p><blockquote><em>establish priority without revealing the work.</em></blockquote><p>The important conceptual distinction is this:</p><h3>You do not need to publish mathematics in order to publish a commitment to mathematics.</h3><p>You can publish or independently timestamp:</p><pre>bd75519bbc5c3a7f...</pre><p>while keeping completely secret:</p><pre>the theorem<br>the proof<br>the Lean code<br>the failed approaches<br>the notes<br>the collaborator discussions</pre><p>Later, if necessary, you reveal the work and show that it produces exactly the previously timestamped commitment.</p><h3>A cryptographically sealed research notebook</h3><p>The end result looks like this:</p><pre>1. Git<br>   exact research history</pre><pre>2. Contributor signatures<br>   who made each contribution</pre><pre>3. RFC 3161<br>   when each state existed</pre><pre>4. End-to-end encrypted Git<br>   who could see the work</pre><pre>5. Optional Bitcoin anchoring<br>   decentralized long-term evidence</pre><p>For the mathematician, however, the workflow remains:</p><pre>git add .<br>git commit -m &quot;Close final induction step&quot;<br>git push</pre><p>That is the important part.</p><p>The strongest security systems are often the ones users barely notice.</p><p>For mathematics, this creates something we have never really had in a standardized form:</p><blockquote><strong><em>a cryptographically sealed research notebook — private while the work is being developed, collaborative when necessary, and independently verifiable when it finally matters.</em></strong></blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=45d2e4d500ff" width="1" height="1" alt=""><hr><p><a href="https://medium.com/openspur/how-to-cryptographically-protect-unpublished-mathematical-work-45d2e4d500ff">How to Cryptographically Protect Unpublished Mathematical Work</a> was originally published in <a href="https://medium.com/openspur">OpenSpur</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Navier-Stokes: Hold on to your papers]]></title>
            <link>https://medium.com/openspur/navier-stokes-hold-on-to-your-papers-7f4e9bcffe61?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f4e9bcffe61</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[math]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Wed, 09 Sep 2026 06:18:51 GMT</pubDate>
            <atom:updated>2026-09-09T06:18:51.220Z</atom:updated>
            <content:encoded><![CDATA[<p>Big tech, at it again.</p><p>Both can’t be true at the same time: that your knowledge work is worthless AND that OpenAi spent 10M USD to prove the Navier-Stokes equation while still leaning on decades of work and allegedly a novel approach built by 2 researchers over the course of a year.</p><p>But in order to get the public’s money in the next block-buster IPO, proprietary AI needs you to believe this. And the PR machine is somewhat succeeding, with big names in science questioning the worth of their post-AI abilities.</p><p>A lot can be said to the very contrary of this. AI can’t draw a stable taxonomy, and it’s why your AI agent memory is fuzzy. It’s a minimization problem that lends itself to the quantum nature of our brains. Novel concepts rest on the taxonomy existing — otherwise, how do you know it’s really novel? First-principle thinking is deeply human, and it’s argueably what extends the horizion of knowledge. And the ever extending horizon of knowledge is what fuels capitalism.</p><p>In terms of knowledge creation, what these AI companies are proposing is a race to the bottom. Remove the incentives, and the expanding of knowledge disappears. Fear sets in.</p><p>What OpenSpur is proposing is the complete opposite: fair rewards for every restriction of the search space every contribution creates. Funds that flow from the public towards computation get a share of the piece of knowledge. The joy of pursuing excelence, novel knowledge and the deeply Godly feeling of being a creator.</p><p>With a grain of belief, the choice is obvious.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7f4e9bcffe61" width="1" height="1" alt=""><hr><p><a href="https://medium.com/openspur/navier-stokes-hold-on-to-your-papers-7f4e9bcffe61">Navier-Stokes: Hold on to your papers</a> was originally published in <a href="https://medium.com/openspur">OpenSpur</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI took our jobs?]]></title>
            <link>https://medium.com/@mihairaulea/ai-took-our-jobs-ca52d9aa7893?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/ca52d9aa7893</guid>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Tue, 22 Jul 2025 07:32:05 GMT</pubDate>
            <atom:updated>2025-07-22T07:32:05.302Z</atom:updated>
            <content:encoded><![CDATA[<p>Knowledge is infinite — AI is not.</p><p>Science is constantly breaking down reality into smaller and smaller components, rewriting the rules as new additions join the science corpus.</p><p>When knowledge about the world changes, knowledge that was built on top of and interacted with deprecated knowledge crumbles — it needs to be rewritten taking into account the new status quo.</p><p>Entropy is the amount of disorder the system is in. Information entropy lets us measure the amount of information new knowledge brings.</p><p>Observing the world and formulating new foundational views is infinite from an information entropy standpoint. A machine with finite resources will never be able to do it — it needs human guidance.</p><p>Hence, generating truly novel knowledge based off a spark of genius, sent from above is not of this world, and not mechanical. And it so happens that it’s an infinite game, as is the horizon of human knowledge. Ever expanding.</p><p>Nobody can take this away. For those of strong faith, new and better things are coming.</p><p>God is great and AI is just another cool tool.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ca52d9aa7893" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Blindspots + broken mechanics + tribes = crash]]></title>
            <link>https://medium.com/@mihairaulea/blindspots-broken-mechanics-tribes-crash-c066ba9d7174?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/c066ba9d7174</guid>
            <category><![CDATA[dao]]></category>
            <category><![CDATA[decentralized-autonomous]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[crypto]]></category>
            <category><![CDATA[decentralized]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Mon, 08 Aug 2022 13:06:53 GMT</pubDate>
            <atom:updated>2022-08-08T13:06:53.870Z</atom:updated>
            <content:encoded><![CDATA[<h3>What does a remote-first Fortune 500 company in corporate America have in common with the EVM?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CIoOFzGvef1La1cCDUwdVQ.jpeg" /></figure><p><strong>Chapter 1 — Tech companies are unexpectedly inefficient</strong></p><p>Crossover has one of the most innovative company playbooks I have ever seen. They have a 1 billion USD budget to buy companies, and they buy one company per week. They have very specialized teams that need to improve on a weekly basis by consuming new technology.</p><p>We were tasked with implementing a new authentication mechanism in all the company&#39;s projects. As I checked on all the teams’ progress, I realized everybody was going through the same learning curve. Yet, after learning the arbitrary config options and playing around with the API, it would have quickly become a 2 hours job for any one of the developers. It ended up costing each team around 40 working hours and tons of EMOTIONAL DAMAGE.</p><p>Why did we duplicate all the work and pain, especially since we were in no hurry? And this stage, the idea was starting to form, but in a very fuzzy way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/851/1*gEIVeSn8BPe3iq3SpoUt5g.png" /><figcaption>How it should be</figcaption></figure><p><strong>Lesson: blindspots are ubiquitous.</strong></p><p><strong>You: Have you ever been in a position where you had to work on something you absolutely dreaded?</strong></p><p><strong>Chapter 2 — Ethereum foundation</strong></p><p>It’s 2018 and i am in the middle of a 1.75M USD due diligence with the Ethereum foundation. Reading through the EVM source code, i see something strange. It had the potential to cut gas costs by orders of magnitude.</p><p>Let me explain. Solidity, the programming language used to write smart contracts gets compiled into bytecode for the EVM(the virtual machine doing the execution). Each heap layer in the EVM is a blockchain-backed consensus entry. The algorithm used to compile Solidity into EVM-readable code was a peep-hole optimization algorithm(highly inefficient).</p><p>So, gas cost could be a lot cheaper by implementing a symbol solver-based optimization of the bytecode output. Less code to go through, lower gas costs. At that time(2018), and still to this day, everybody was complaining about Eth gas costs. I emailed the Ehtereum foundation about this. I even wrote on a github thread about it. They posted a gitcoin bounty worth 500 USD and tagged me.</p><p>This problem was worth way more — judging by the daily gas costs. I decided to drop the issue altogether.</p><p>I was beginning to realize this is an endemic problem. How can we layout fair incentives and let the community self-organize around problems to be solved instead of taking other people’s money and giving them to our friends as grants?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/691/1*2YdoTWSsNz8yP1x-mbU1hw.png" /><figcaption>Know exactly what you’re getting into, and how much it pays</figcaption></figure><p><strong>Lesson: There is no fair remuneration system. No way to capture cognition from willing participants. This costs everybody.</strong></p><p><strong>You: Have you ever been in a position where you could better significantly one of the products you were using, would have loved to but had no incentives to do so?</strong></p><p><strong>Chapter 3 — Dashboard hero</strong></p><p>If even the largest organizations on Earth have this blind spot, it’s time to start experimenting with a framework that deploys skills like a Tetris game — get the perfect piece to match the current situation. Sometimes, you can do things in parallel. Some tasks are not primitively recursive, some are. Sometimes the tasks draw a DAG, sometimes it is a tree. More on this in a separate article, though.</p><p>Tested a hypothesis: are startups interested in having a 3rd party build their dashboard? It turns out, they do. For the next 6 months, I built a reliable framework and playbook for building dashboards, with authentication and billing mechanics(and a marketing funnel that still works to this day).</p><p>Information and experience were harnessed from around 50 consultants and developers that actively contributed to both the playbook and did parts of the actual work. I now had a factory that turned dashboard specifications into working software.</p><p><strong>Lesson: One hour from the right expert at the right time can save 100s of hours of guesswork.</strong></p><p><strong>You: Have you been part of a team that was derailed from their core concerns to work on something lateral to the team’s mission?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/132/1*ykJu5PCoDWxD9hhs_cxRFw.png" /><figcaption>DashboardHero logo</figcaption></figure><p><strong>Chapter 4 — Codebuddy</strong></p><p>How do we scale this process from a task-specific one to the whole industry? Codebuddy was the 1st attempt at turning the task-to-developer(s) matching algorithm into a product.</p><p>We built the IDE plugin and pitched it to GM, Crossover, and a couple of others.</p><p>We quickly noticed a couple of things:</p><ol><li>Not enough skill inside companies for all the would-be-problems(separate article). On the other hand, introducing the notion of external consultants that employees can use at discretion was not looked fondly upon.</li><li>Why would i help you? This question came up way more than you would think. And, indeed, even if you’re my colleague, why would I prolong my workday to help you?</li></ol><p>Given that we had already indexed 1M developers, about which we knew exactly what they shine at, we could have pivoted into an HR startup. I started this journey to offer freedom to as many individuals as possible, not to annoy people with job offers.</p><p>We called it quits. Several innovative algorithms and a lot of bright ideas had come up — but a cohesive blueprint for reliably and repeatedly liberating individuals from their 9-to-5s remained elusive.</p><p>The discover/build/test phase was still in full swing. Don’t act. Just think. For now.</p><p><strong>Lesson: We need fair, transparent incentives. And a brutal divorce from salaries.</strong></p><p><strong>You: Have you ever wanted to pick somebody’s brain for a couple of minutes? Would this have spared you hours of frustration?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*JrlNtVewb-F-o5o7RY30SQ.gif" /></figure><p><strong>Chapter 5 — the open-source ecosystem</strong></p><p>I took a look at how far the rabbit hole goes — and will publish what I call Openomics in a separate article, along with the data I used for the study.</p><p>The average software product depends on 98 other open-source projects. What’s the relationship between the attention a project gets, and the number of issues that get opened on it? Preliminary data suggests it’s a linear relationship. The more popular the project is, the more issues are opened.</p><p>The average resolve time for an issue on Github is 28 days. The average resolution on Stackoverflow is 3 days.</p><p>The notorious problem of open-source funding is a stupid one. It stems from greed.</p><p>The most precious resource is time. If we take into account all the time lost globally with badly configured projects, integration and deployment problems, lack of documentation — reflected in all the opened issues on Stackoverflow and Github — without mentioning the forums, and the issues that never make it to the light of day, the number is stupendous.</p><p>How much did the Log4j vulnerability cost? How much did it cost Elasticsearch the fact that AWS downright stole their intellectual property? And how much did it cost in terms of lost trust?</p><p>We are interconnected. We depend on each other’s happiness. Pull a thread here and find that it’s attached to the rest of the world.</p><p>How can we build a fair, transparent system that enables innovators to associate and create with minimal friction? In this system, every contribution is remunerated, so that, when you use or deploy a package, you’re pretty sure the people working on it were very, very happy. And so will you.</p><p>Moreover, if you do run into trouble, maybe you can summon someone that’s better than you at solving it. And you, in turn, can solve other people’s problems. A defragmented organization with perpetual rewards and no overlords.</p><p>The pitch for this system was very well received. In the meantime, all my consulting projects were being built using these principles.</p><p><strong>Lesson: Building in the open outpaces every company on earth. We are already deeply connected. Somebody needs to make it obvious, and start assigning and paying transparent and fair incentives.</strong></p><p><strong>You: How much time have you wasted this week with arbitrary framework configs or a lack of full interoperability?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/543/1*nLbmRytOXORmaWqfwn1RlA.jpeg" /></figure><p>With OpenSpur, we took all the lessons above and wrapped them in an on-chain product. We’ve tested the product, and delivered software with it. It brings a 2–3x increase in pay to the developers involved and a speedup of 5x, on average.</p><p>We’re currently getting ready to showcase the product and onboard partners. Yes, this is exactly the right time, as the market is now dead silent.</p><p>We thought deeply about it. It’s time to act.</p><p><strong>LFG!</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c066ba9d7174" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Breaking tech monopolies]]></title>
            <link>https://medium.com/openspur/breaking-tech-monopolies-295367f619e0?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/295367f619e0</guid>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Mon, 01 Nov 2021 13:19:32 GMT</pubDate>
            <atom:updated>2021-11-05T09:04:52.053Z</atom:updated>
            <content:encoded><![CDATA[<h3>Moving towards decentralized software development</h3><p>It’s easy to extract value from your product if 10k people pay 10 USD per month.</p><p>It’s really hard to extract value from a product if 1M people would pay 10 cents per month to use it(payment gateway fees, customer attention).</p><p>And it’s almost impossible to break into an established supply chain — no matter how superior your solution is.</p><p>Innovation is built from a lot of 10 cent ideas. So although your opensource project might be worth more than all the companies that use it combined, you are left with little options for monetization.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*I0MOJSf92Sk4S_4BoDz1Aw.jpeg" /></figure><p>Given that email was first built by developers and for developers, and is now ubiquitous — can a new open-source monetization mechanism create the precedent and the blueprint for a better way of working on what you love, what the world needs, what you’re good at, AND being rewarded fairly for it?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/625/1*ZzB2Zrw7LLFxC21cvYOfEQ.png" /></figure><p><strong>The proposal</strong></p><p><strong>Turn code into electricity. </strong>Into a pay-per-use commodity. Connect the money hose directly to creators, bypassing the gatekeepers. Obtain a positive feedback loop by transparently and fairly paying every project contributor.</p><ol><li>Flatten the hierarchy and extract cognition optimally from all participants. Instead of top-down, go bottom-up; an A.I. recommender system advertises work to be done to the best-fit developers. Developers choose what they work on, based on what they’re good at, what they care about and what gets them paid the most.</li><li>Reduce friction and let developers associate and create together freely. No contracts, no gate-keeping. If the work is good, it gets remunerated by a smart contract.</li><li>Defragment organizations. Support means trading 20 hours of work from someone who isn’t an expert in a very specific problem(but knows the domain language of the task) for one hour from an expert.</li></ol><p><strong>How to start?</strong></p><p>For simplicity, start remunerating open source projects and let authors and maintainers figure out where they allocate part of the funds they receive.</p><p>In this new model, all open source project authors are remunerated based on the impact they have on the network of open source projects.</p><p>But how do we start injecting capital into this network, so we get the (positive feedback)wheels turning?</p><p><strong>Capital injection</strong></p><p>OpenSpur will be rolling out the first decentralized managed database cloud solution. There are 900 database systems in existence, each with unique strengths and tradeoffs. The first deployment will contain about 100 database offerings — more than any other centralized counterpart(Heroku has about 25 database offerings in their add-on marketplace).</p><p><a href="https://www.singlestore.com/">SingleStore</a> and <a href="https://aiven.io/">Aiven</a> are doing this in a traditional, gated system. Can a decentralized network of developers working on a common goal offer a better product?</p><p><strong>The prototype</strong></p><p><strong>Networks and network effects. </strong>Let’s see how to leverage them.</p><ol><li>Maybe Spring Data doesn’t have bindings for your database tech yet — but directing 5% of your profit to Spring developers will bring it to market. This will increase adoption of your tech and increase the revenue you get. Rinse and repeat with another ORM or community tool.</li><li>Common hardware infrastructure for all projects. The platform leases and releases cloud machines as demand fluctuates. The project authors are oblivious to this — it’s part of the platform.</li><li>One transparent, low-fee payment gateway. 95% of the profit goes to the project owners. 3% goes to the rest of the open source community, in orders of magnitude proportional to the contribution(think transitive dependencies and ecosystem projects such as ORMs) to the project bringing in the revenue. The payment token also captures the value of the network .</li><li>One unified software interface to store metrics, logs and handle user accounts. One UI for all projects.</li><li>One common marketplace for all cognition. Be it support, architectural consulting or work. All of it with perpetual rewards. As a project owner, you get a cut of everything.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F7ahR0sTxZx1mgyrPlL8Pg.png" /><figcaption>The OpenSpur dashboard</figcaption></figure><p>To participate, an open-source database developer simply writes a recipe to deploy the technology, and sets the hourly running price. Integrating with the logging, metrics and support feature(which will be paid) is simply a matter of performing a set of https requests.</p><p>This would validate the hypothesis of <a href="https://en.wikipedia.org/wiki/Decentralized_autonomous_organization">distributed autonomous organizations</a>. And, in a larger sense, the value and importance of freedom — with a quantifiable, economic bottom line. We believe in “free”(not of charge) and “open” — let’s make it happen!</p><p><strong>Closing thoughts</strong></p><p><a href="https://openspur.com/">OpenSpur</a> will be coming out of stealth soon — we’re currently onboarding select projects. Enabling developers to associate and create with minimal friction — and being remunerated for it. If you are a project owner and this is a worthy mission for you, and something you want to be a part of, <a href="https://gitter.im/OpenSpurAlliance/community">join our community</a> and <a href="https://twitter.com/OpenSpur">follow us on Twitter</a> for updates.</p><p>Excelsior!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=295367f619e0" width="1" height="1" alt=""><hr><p><a href="https://medium.com/openspur/breaking-tech-monopolies-295367f619e0">Breaking tech monopolies</a> was originally published in <a href="https://medium.com/openspur">OpenSpur</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to build your SaaS’s dashboard]]></title>
            <link>https://medium.com/hackernoon/how-to-build-your-saass-dashboard-5e3d40dd2003?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/5e3d40dd2003</guid>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[dashboard]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[saas-dashboard]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Wed, 22 May 2019 07:54:41 GMT</pubDate>
            <atom:updated>2019-07-12T06:51:45.141Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to build your SaaS dashboard</h3><h4>Five steps you need to follow through — this ain’t rocket science, but it could break your product.</h4><p>I have failed more businesses that i can count — but my last one. Like Warren Buffet says, you only have to be right once. In the process, i have learned to ship quickly and fall out of love with my ideas and tools. Here’s a short list of tools you could use to ship your SaaS dashboard fast and effectively.</p><h4>Situation</h4><p>You or your core team have come up with a novel algorithm, have scraped and labeled some data, etc. You think others would pay to have access to this. Not focusing on validating this hypothesis as soon as possible will kill your game. Reinventing the wheel is at the fore-front of failed products and broken dreams.</p><p>No, you’re not the master puppeteer— you can’t build a responsive UI from scratch. No, you’re not a data architect — you won’t select the killer database for your use-case just by thinking about it.</p><p>You’re just a guy who has a dream. Learn to accept that, let go of your need of control, and delegate heavily. When you try to go for custom in areas that are not part of your core business, a little Pareto(principle — the idea that 80% of your wins come from 20% of your efforts) starts crying.</p><p>In a healthy body, the brain does not tell the liver, heart and lungs what to do. Accept that there are other people/tools who can do better than you in certain areas, learn to delegate, and do it often.</p><p>Let’s look at a couple of things you’ll need to implement to extract value from your core product. These include implementing the UI, user management, handling data and access to it and finally putting everything together(architecture). Of course, every modern application should be able to ship changes fast, so we will be looking at how to implement a CI/CD pipeline. I have jumped through all this hoops in less than a week, but it took me quite some time to put the process in place. Enjoy!</p><h4>Responsive UI/UX</h4><p><strong>Problem</strong></p><p>What does playing with bootstrap have to do with the data you are selling to your customers? Absolutely nada, yet here are your engineers burning through countless hours to nail that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*8I5Rgox546aVlN-MsSZAzQ.gif" /></figure><p><strong>Typeform</strong> can’t even serve mobile users — that’s 58% of the users browsing the Internet, atm(<a href="https://www.stonetemple.com/mobile-vs-desktop-usage-study/">source</a>). They’ve been around since 2012.</p><p><strong>Solution</strong></p><p>As stated above, stop re-inventing the wheel. Go for something that works decently, out of the box. There are a ton of free solutions. I used <a href="https://github.com/puikinsh/Adminator-admin-dashboard">Adminator</a> in my projects, but Google might enlighten you further.</p><p>It has all you could dream of to mock and build a responsive admin dashboard, like the one your SaaS needs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LHmHBptVlEv_anJQ2lQ92Q.png" /><figcaption>Looks decent — does it’s job. You’re not marrying it. Go for it today!</figcaption></figure><p>When mocking the interface, you may want to push to a remote server so that members of the team discuss it. Adminator uses webpack — you can export the build of your mocked interface, and use an express server plus a heroku deployment to serve the build to the world! Now you’ve got a CI/CD pipeline for your mock!</p><h4>User management</h4><p><strong>Problem</strong></p><p>What does any of the dancing around user accounts have to do with the core of your startup?!? Here’s a a major fu**up: index.co, made by The Next Web. The reset password does not work, neither do the OAuth integrations. Would have expected better from a product charging 25 USD per month.</p><p>The 1st website that had an account feature probably came out 20 years ago. This shouldn’t be that hard, right? Ask Auth0, who has just attained unicorn status offering a solution for that. And it still doesn’t work out of the box.</p><p><strong>Solution</strong></p><p>I have found Firebase to be of immense help here. It handles everything you would expect of it. Register and sign in with a number of methods(classic email/password, Google/Fb account, etc.). Verify email address feature. Phone number verification. Forgot password.</p><p>It even comes with a back-end user management solution, so you can disable or delete accounts. It could also be helpful if you wanted to manually approve users — just store a variable on the user account, and default it to “notApproved”.</p><p>Another use case that i have faced is collecting some data from the user before registering. I have solved this by linking to a Typeform survey, and using a webhook when a form gets completed.</p><p><strong>Data storage, retrieval and access control</strong></p><p><strong>Problem</strong></p><p>Asking yourself “Will my storage solution scale?” when you barely got your first 20 customers are words that shouldn’t be uttered. What will kill your SaaS startup is the high cost of customer acquisition and low life-time value or high churn rates. I say OR because one is enough.</p><p>Forget ideology, let’s talk practice for a minute. To put this in perspective — what are the biggest nosql deployments that you know of? Fifa has some 250 servers running to persist data in MongoDB clusters(<a href="https://www.mongodb.com/mongodb-scale">source</a>). That’s ONLY 250 servers. Clusters of about 100 instances are normal among large deployments(<a href="https://www.youtube.com/watch?v=BfPDZf2wmqg">source</a>).</p><p>I admit —seldom, the questions you have to ask impose a unique kind of storage technology. When your business picks up steam, you could go from 125 db instances, to 48 and finally to only 3 — by playing around and selecting the right tech(see Adobe Behance’s use case <a href="https://www.youtube.com/watch?v=BfPDZf2wmqg">source</a>). That’s important, because at scale, the bills pile up quickly. But you’re not there, so you shouldn’t worry.</p><p>What you want is a quick way of connecting the data to the UI and controlling access to it.</p><p><strong>Solution</strong></p><p>Again, i think Firebase is immensely helpful. You don’t get to manage database instances, they’re already there. Also, you don’t need to concern yourself with connecting to the said data from front-end, and security is well taken care of for you. Like MongoDB, it’s document-based. You’ll be able to manipulate the data from javascript, using the firebase library, in just a couple of lines.</p><p>Rules for accessing the data are available. Important inflexion points: <strong>authenticated</strong> vs <strong>non-authenticated</strong> users and <strong>read</strong> vs <strong>write</strong>. You can also group users and revoke access to a certain document in your db. Or, you can explicity allow just one user access to a certain document. Rules to automatically set this up can be written(so you don’t end up manually pairing users to documents).</p><p>A chron job can be used to push fresh data to Firebase. Every time a user makes a call to retrieve data, you can count that and charge accordingly.</p><p><strong>To sum this up,</strong> you want to:</p><ul><li>push your newly acquired data regularly</li><li>store and retrieve data</li><li>display it in the UI</li><li>control who gets to have access to it</li><li>charge based on usage</li></ul><p>All this can be done in a couple of days, if you know what you’re doing.</p><p><strong>Architecture</strong></p><p>Okay, so by now you’ve got a couple of scattered html, css and js files. You’ve got a web app with no live data — this is the blueprint for your dashboard. User management is there. You also have a pipeline to push the data that your users will want to consume. You’ve got the data manipulation right. How to put everything together?</p><p>I advise you go for a feature-first React architecture, with Redux for the in-memory data store. Here’s how this would look.</p><p>React separates each visual element into it’s own class/file. You want this, because you want to stick with the DRY principle. In adminator, the menu and top bar will be common to all the elements in your page. You’ll want to be able to use those without pasting the code each time.</p><p>You’ll want to handle state. I love redux — state changes are only made through pure functions, so everything is nice and predictable. There’s a two-way binding in place between your store and your React components.</p><p>Things can go in two directions: either your UI sends a command that changes your data store and Firebase data, or your Firebase data changes, and your store updates, which prompts the UI to update.</p><p>Simple and clean, right?</p><p>What about the feature-first architecture? Here’s a pretty descriptive picture:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3a0pzW6SAfnJHHhcyxMXfQ.png" /></figure><p>Let’s say you want to display a chart in your React app — the chart has a visual component, and a data component. In a feature-first architecture, the only files you would need to touch would be under the Components/CustomChart folder. This makes development easy and scalable(you can have any number of engineers making changes to your UI). There are numerous other benefits to this, and i advise you to start reading a bit on the topic.</p><p><strong>Infrastructure(CI/CD, quick deploys)</strong></p><p>Now that you’ve put everything together, it’s time to deploy your solution. You will want to have a staging environment — a place to manually test before it goes out into the real world. And you want a production environment, to which you push after you’ve smoked tested your changes. You can achieve this by creating two separate firebase projects, and switching necessary for each by switching credentials based on a variable.</p><p>Also, it would be nice to have a way of automatically pushing changes. You can come up with a script in your npm that builds the solution, navigates to the build folder and does <strong>firebase deploy</strong>.</p><p><strong>Conclusions</strong></p><p>Although very high-level, we’ve jumped through all the hoops necessary to launch your awesome SaaS to the world. I challenge you to do this in the next 2 weeks(it is possible).</p><p>As always, speed is key — you need to test, improve and operate the funnel, to validate your hypothesis. Is it really the case that you are building a killer product? Is the customer acquisition cost low enough, and the value you are generating for your users, high enough, to have a viable business?</p><p>The sooner you know, the better. If you’ve just finished building your core product and want to start exploiting it fast and effective, here’s my <a href="https://dashboardhero.dev/">dashboard for SaaS product</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5e3d40dd2003" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hackernoon/how-to-build-your-saass-dashboard-5e3d40dd2003">How to build your SaaS’s dashboard</a> was originally published in <a href="https://medium.com/hackernoon">HackerNoon.com</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A deep dive into deep tech — for execs]]></title>
            <link>https://medium.com/neo4j/a-deep-dive-into-deep-tech-for-execs-1dc22b686915?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/1dc22b686915</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[neo4j]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[database]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Fri, 01 Mar 2019 20:34:27 GMT</pubDate>
            <atom:updated>2019-03-01T20:34:27.119Z</atom:updated>
            <content:encoded><![CDATA[<h3>A deep dive into deep tech — for execs</h3><p>Graph databases are powerful because they allow for the exploration and analysis of the connections in the data.</p><p>For SQL, it’s impossible to answer in real-time to a query with 5–7 JOINS. Document-based databases(think Mongo and Couch) lock in place the connections between the data; once you establish a relationship between the objects(nesting), you would have to go through 2 entire datasets to uncover a new relationship between 2 objects.</p><p>For graphs, following an edge(relationship) comes with almost zero cost, which opens up a world of possibilities.</p><p><a href="https://medium.com/@mihairaulea/why-deep-tech-products-fail-239d142cf1f1">Deep tech is hard to explain</a>, and graph databases are no exception.</p><p>Everything in tech has a language of its own and for good reason. Have you tried having a conversation in a foreign language, with a native, when all you knew were 200 words?</p><p>Explaining deep tech is more than that; it’s not only that one can’t use a word that abstracts what would be put in tens or hundreds of other words; it’s also the fact that the implications and ramifications of one concept are huge. After all, every programmer uses a maximum of 20 or so basic language constructs (declaring variables, executing an instruction several times, based on a condition, etc)— but look at how diverse the tech world is.</p><p>Every flavor of storing data out there does so by using rows and columns; something quite similar to Excel spreadsheets. A paradigm that makes relationships between data as important as the data itself exists and the implications are huge.</p><p>Without graph databases, we wouldn’t have Google; its search algorithm runs on top of a graph. Facebook uses a graph internally to store the users and the relationships between them and it’s of paramount importance when suggesting new friends and events. But what is a graph?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*Tf52dcEgx1FlkpRRHMEaAg.png" /></figure><p>There you go! That’s a graph. Can you draw that?</p><p>It labels the nodes, from 1 to 6. It tells us that 1 is connected to 2 and 5. That 4 is connected to 6. That’s it. Thank you!</p><p>I don’t know the vertical you’re in; but let’s assume you know and use social media(Facebook, LinkedIn, etc).</p><p>Let’s have a look at how you would model a network of people based on who knows who.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*7CRFe7dPdWgqezMrJ73uAQ.jpeg" /><figcaption>User 4 knows User 5. User 7 knows user 6. Ignore the arrow direction, for now.</figcaption></figure><p>Here’s how one would model a network of friends that can author posts, and like each other’s posts. I bet you can now read it on your own. Users are not labeled, in the below. Take the User in the center; how many friends does he have? How many posts has he written, and how many has he liked?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3UImThl3F3yxbPRKqbsbig.png" /></figure><p>Now, on to some serious business. Looking at the network, who would you contact to post something about your business? Which of the users is most connected to other users?</p><p>If only friends can see each other’s posts, and some users meet in person and show posts from their network to each other, what’s the probability of two unconnected users seeing each other’s posts? Just joking.</p><p>As stated earlier, the implications of this easy to understand tech are remarkable. I won’t bore you with my work on the core of Neo4j or satellite projects — but here are some high-level examples of how I brought ROI as a consultant and even made the use-cases possible.</p><p><strong>Case study 1 — Adventia Prevention(bioinformatics)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C4UGo_y6LOtnVhI4tj19Gg.gif" /></figure><p>Adventia is a personalized medicine tool that suggests treatments. Doctors use it to make better and more informed decisions for their patients.</p><p>Each doctor has a number of patients, and each patient has blood work data, a history of previous disease and physical attributes(such as age, height, weight). Among other things I did for Adventia(built an interpreter and execution engine for a flavor of a programming language that (even) doctors can write, implemented Bayes probabilities on top of the graph), i made sure their queries were possible(SQL systems wouldn’t support them because of the huge amount of data that needed linking together) and enabled them to ship on time(writing SQL queries for our types of questions would have been too specific, and required several individual queries where graphs could do it with one, more abstract query).</p><p>All in all, it took 3 months to reshape the future of Adventia. In the end, I made some of their very needed use cases, possible and trained their team to use graph databases — after all this, their team velocity increased. ROI for the win!</p><p><strong>Case study 2 — Atlas, or how to put the world map in a graph</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZleJwm0MbskYtpAB1F5vpg.png" /></figure><p>Have you ever wondered what a graph of all the roads in the world would look like? What kind of amazing insights one would be able to uncover if such a graph would be put in a <a href="https://neo4j.com/developer">Neo4j</a> database? Amazing information can also be unveiled if one could extract the most connected cities in Europe, and it would definitely come in handy for a vacation planning platform.</p><p>I have published a dedicated blog post about this with the leading graph database provider, Neo4j. Find it <a href="https://medium.com/neo4j/how-i-put-the-world-map-in-a-graph-422b651780e9">here</a>.</p><p><strong>Case study 3 — Blue Air, and finding (definite)similarities in web pages</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/972/1*Ix7hYmPS7AN_XvqoQxQfyg.png" /></figure><p>Have you ever looked at how HTML looks? It’s a hierarchical system. My client was employing AI to determine if too web pages are alike, by looking at the source code. The problem? AI works with a degree of uncertainty(heuristics). Graphs can yield deterministic answers.</p><p>Not only did my client save on computing power, but the system ran smoother after the upgrade, with zero false positives.</p><p>It seems this is a trend to follow; in his book, <a href="http://www.connectedthebook.com/">Connected</a>, Dr. James Fowler argues studying relationships is more important than the data contained in the nodes.</p><p><strong>Who else is using graphs?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*GBJagrYNtSwc-6Ytr4nolg.png" /></figure><p><strong>Conclusion</strong></p><p>I believe there is a huge opportunity in the graph space, and there is a lot of value to be created by educating yourself about the space and using graph databases. If you read so far and want to know more, I would be happy to discuss how it could be used in your company. <a href="https://calendly.com/mihairaulea">Book a 15-minute call now!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1dc22b686915" width="1" height="1" alt=""><hr><p><a href="https://medium.com/neo4j/a-deep-dive-into-deep-tech-for-execs-1dc22b686915">A deep dive into deep tech — for execs</a> was originally published in <a href="https://medium.com/neo4j">Neo4j Developer Blog</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why deep tech products fail]]></title>
            <link>https://medium.com/hackernoon/why-deep-tech-products-fail-239d142cf1f1?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/239d142cf1f1</guid>
            <category><![CDATA[neo4j]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[graph-database]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[business]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Tue, 19 Feb 2019 14:11:56 GMT</pubDate>
            <atom:updated>2019-02-19T21:33:41.032Z</atom:updated>
            <content:encoded><![CDATA[<p>It’s fascinating how the life cycle of a product looks. My unpopular opinion is that communication is the biggest killer of deep tech products.</p><p>In Crossing the Chasm, <strong>Geoffrey Moore </strong>identifies 5 steps every innovative company goes through. First, 2.5% of the market adopts the product. This first batch of users are innovators themselves — they understand the tech well. Then come the early adopters, that comprise roughly 12.5% of the market — they are somewhat curios, but don’t get everything.</p><p>The gap between early adopters and early majority, is called the chasm. How fast the early majority gets acquired(or not) as customers is crucial, and makes or breaks a product.</p><p>This is true for the blockchain, graph databases and A.I.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*W1k_p1k5TuDVv_gTWotRYg.jpeg" /></figure><p>In Zero to One, Peter Thiel differentiates between <strong>horizontal</strong>(polynomial) and <strong>vertical</strong>(exponential) progress.</p><p>Taking things that work somewhere and making them work everywhere is an example of horizontal progress. China is one example; its 20-year plan is to become like the United States is today. If you know your food ordering apps: GrubHub is the US leader and first mover while FoodPanda dominates the markets in Asia. Compare the valuation of these 2 companies, and you’ll understand that merely copying others will bring you less profit.</p><p>The single word for vertical, 0 to 1 progress is technology . Properly understood, any new and better way of doing things is technology — and deep tech is in most cases, the fuel.</p><p>My unpopular opinion is that many deep tech products don’t cross the chasm, because communicating their strengths and return on investment is impossible. The product can be communicated clearly to other peers(innovators), and some of the early adopters. Then, game over. This is ultimately a big loss, because deep tech is what drives vertical progress.</p><p>The communication chasm</p><ol><li>Developers can communicate with other developers</li><li>Business execs, marketers can communicate with their peers</li><li>There is no common language for business execs and developers</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*sW__mw2kvTgy4m7H0gzcZw.jpeg" /><figcaption>The elephant in the room</figcaption></figure><p>Let’s have a look at my favorite deep tech product, graph databases and the company behind the most popular solution, Neo4j.</p><p>If you’ll have a look at the Neo4j blog, you’ll quickly understand. No one article addresses business executives properly. It’s deep tech people, talking to other deep tech people. Some articles try to be friendly to the business guy, but end up exploding with tech jargon and references.</p><p>My best stab at explaining graph databases in a sentence is this: instead of storing data in a tabular format(Excell), we store it in a easy-to-draw, whiteboard friendly manner that puts an emphasis on relationships in the data, instead of columns and rows. The implications to this are monumental — it’s what made the Google search empire possible. But the tech continues to be heavily overlooked(even inside Google, despite push from C-level and V-level people).</p><p>Here’s a bonus unpopular claim: i think *most* of the code(think Pareto distributions) that covers 80% of the needs out there has been already written.</p><p>You are a developer, and want to do deep tech — it’s probably the most rewarding intellectual challenge out there. The challenge(and bulk of the work) in the coming years will be to correctly communicate and identify needs based on the product, and some clever, business-oriented, integration work — not implementing it.</p><p>Companies, business execs — if you want your business to create genuinely *disruptive*(cringe!) products, pay attention to and be patient with deep tech consultants and developers. If you want to transform what you can do with your data, <a href="https://mihairaulea.github.io/GraphConsultingServices/">hire me</a> or <a href="https://calendly.com/mihairaulea">book a free 15-min blitz call</a>.</p><p>Happy 0-to-1-ing!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=239d142cf1f1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hackernoon/why-deep-tech-products-fail-239d142cf1f1">Why deep tech products fail</a> was originally published in <a href="https://medium.com/hackernoon">HackerNoon.com</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From zero to graph hero, in 5 days and 5 steps]]></title>
            <link>https://medium.com/hackernoon/from-zero-to-graph-hero-in-5-days-and-5-steps-34a1853b75c8?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/34a1853b75c8</guid>
            <category><![CDATA[database]]></category>
            <category><![CDATA[business-intelligence]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[business-development]]></category>
            <category><![CDATA[sql]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Wed, 13 Feb 2019 11:46:01 GMT</pubDate>
            <atom:updated>2019-02-13T11:46:01.996Z</atom:updated>
            <content:encoded><![CDATA[<p>Sharing my playbook for graph database projects. Will be happy to answer any questions or offer advice.</p><p>Ask Siri who is the quarterback of the New York Jets and you get a list of the current quarterbacks right away. Ask Siri who is Taylor Swift’s boyfriend, and it will know.</p><p>Ask who are the members of the Irish band, The Chieftains, and Siri doesn’t know. Even if you ask about who are the members of the Irish Band U2, it doesn’t know. It can provide you information about Bono, but doesn’t have an idea of the members of U2.</p><p>However, Google Assistant know to answer the questions above. Can you guess what happened?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/1*XUbv1vE6qKHhUJDuyZbluA.jpeg" /></figure><p>Despite an almost 50% growth YoY of graphs, the tech sees a small fraction of the database market share. Yet it’s power, flexibility and speed of development reamins amongst one of the best. Better yet, it leads the way to use cases that are beyond the grip of any other system i have worked with.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vVxPojYVfrHXlrlfExMw1w.png" /><figcaption>Source: <a href="https://db-engines.com/en/ranking">https://db-engines.com/en/ranking</a></figcaption></figure><p>As an independent consultant for the past 7 years, i have produced a playbook that can help you levrage the power of graphs in 5 days. Think of it as a high-performance factory. Below, you’ll find the needed steps and how to go about them.</p><p>Let’s have a look at how your company or product can levrage this tech. What are the steps you can go through, and what results can be expected at the very end of this process?</p><h4>Overview</h4><ol><li>Import your current data</li><li>Basic queries</li><li>Handle data duplication</li><li>Network Analytics/Advanced graph algos</li><li>Deployment</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XGZa_Z7KwjRxyAqb9qGOSg.png" /></figure><h4>1. Import</h4><p>You already have one or several databases(mysql/postgresql, mongodb, Cassandra, etc.). The goal of this stage is to explore the current structure of your data and translate it into a graph. Part of this stage is also known as ETL(Extract-Transform-Load).</p><p>If working with an external consultant, make sure to encrypt/obfuscate sensitive data and get an NDA in place, before handing it over.</p><p>Using the right tools, this can happen in under one day. Naturally, actually loading the data in the graph may vary, timewise. Expect to let the system run for about 30 minutes for each 10GB. It’s quite rare that all the data needs to be imported, so good communication before starting the work goes a long way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1RIKJa2oAXLRyuMsZVrC8A.jpeg" /><figcaption>From 2 tables in a relational database, to nodes and relationships in a graph database.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3MmamxpygRzjZ71-cY8xVA.jpeg" /></figure><p><strong>Side-effect bonus</strong>: Non-technical people can understand the structure of the data, as graph databases are very whiteboard friendly.</p><h4>2. Basic queries</h4><p>We need to keep our new data in sync, so we need to connect existing actions that create, update or delete items in your tabular db system to the new graph database.</p><p>In this step, we simply evaluate your current SQL queries, and translate them to Cypher.</p><p>This step usually spurs some very nice Aha moments. What was impossible in the tabular view(doing 5 joins in a row, in real-time) is now highly achievable.</p><p>If we’re talking about 10 queries and 4 pipelines that need to also direct the data to the graph, this can happen in one day. Otherwise, it might vary.</p><p><strong>Testing</strong> can be done by running two systems in parallel. Let’s name the two databases SQL1 and Neo2. Here is how you would test you got this step right:</p><ol><li>Do an import from SQL1 to Neo2.</li><li>Apply a sequence of queries to both systems. Example: Query1, Query2, Query1 to SQL1 and GraphQuery1, GraphQuery2, GraphQuery1 to Neo2.</li><li>Do an import of SQL1 to Neo2. Assert that the 2 databases contain the same data.</li></ol><p><strong>Benchmarking </strong>the two systems will prove which one handles certain queries better. Some functionality previously unattainable with SQL will become possible with Cypher.</p><p><strong>Side-effect bonus</strong>: An understanding of how graphs work. The first new possibilities start to unravel.</p><h4>3. Data duplication</h4><p>Take the queries in step 2, and connect them to your current systems. These could be methods that handle http calls in your microservices, database triggers, etc.</p><p>Be sure to use threads, or some async mechanism, so submission of the data to the databases happens in parallel.</p><p>Hint: if updating Neo4j in real-time is not a priority(using it for analytical capabilities, to present aggregates of the info, recommender system that does not need fresh data, etc) you can use a smaller database instance and use a queue(Kafka, RabbitMQ) to throttle the update/usage.</p><h4>4. Network analytics and advanced graph algorithms</h4><p>I have applied over 200 algorithms on top of graphs. It’s science, yes, but it’s also about having the intuition of what is possible and knowing how to ask the right question(just like in real life).</p><p>In my consulting projects, only once have i had to write a complex algorithm from scratch. Usually, you will find everything you need, under a permissive Open Source license.</p><p>If your engineers know what they’re doing, this can take less than 2 days.</p><p>For an in-depth dive and some intuition on what can be achieved, have a look at the <a href="https://medium.com/@mihairaulea/top-5-applications-of-graph-databases-b186c3105fd1">top 5 interesting applications of graph databases</a>.</p><h4>5. Deployment</h4><p><strong>Deploying to the cloud is a hot topic nowadays, and LinkedIn predicts that the hottest skill for techies this year will be cloud knowledge. </strong>While building the Chinese version, LinkedIn used Neo4j, the leading graph database. Their conclusion was that they were able to move very fast, and didn’t have the need for a database administrator.</p><p>Deployment to a managed cloud service can happen in a couple of hours. Deploying your own pipe can take a couple of days.</p><p>Last tip: while you won’t be needing an administrator, you will be needing a cloud deployment, and i use <a href="https://www.graphenedb.com/">GrapheneDB</a> for most of my projects.</p><p>Good luck, i wish the new functionality you build returns an astronomical ROI. If you want to make sure, <a href="https://mihairaulea.github.io/GraphConsultingServices/">hire me now</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=34a1853b75c8" width="1" height="1" alt=""><hr><p><a href="https://medium.com/hackernoon/from-zero-to-graph-hero-in-5-days-and-5-steps-34a1853b75c8">From zero to graph hero, in 5 days and 5 steps</a> was originally published in <a href="https://medium.com/hackernoon">HackerNoon.com</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Building a playground for top developers]]></title>
            <link>https://medium.com/swlh/building-a-playground-for-top-developers-779d9ee66559?source=rss-4a51aaa081cb------2</link>
            <guid isPermaLink="false">https://medium.com/p/779d9ee66559</guid>
            <category><![CDATA[software]]></category>
            <category><![CDATA[developer]]></category>
            <category><![CDATA[agile]]></category>
            <category><![CDATA[development]]></category>
            <category><![CDATA[software-development]]></category>
            <dc:creator><![CDATA[Mihai Raulea]]></dc:creator>
            <pubDate>Wed, 13 Feb 2019 07:06:00 GMT</pubDate>
            <atom:updated>2019-02-13T07:06:00.787Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: TopDev is a fictional company, something i would like to build at some point.</p><p>There’s a Japanese word that describes what it’s like to fully commit to an activity and immerse yourself in it — <strong>kamiwaza</strong>. Literally translated, it means “like the gods.” That might sound terribly arrogant, but it’s simply not.</p><p>Prolific and controversial writer, Bukowski, talks about this “no other feeling like this” in his poem, All the way.</p><blockquote>“You will be alone with the gods, and the nights will flame with fire”.</blockquote><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fk6_QUhUPrF4%3Ffeature%3Doembed&amp;url=http%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Dk6_QUhUPrF4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fk6_QUhUPrF4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/32aeec4821e2582b1e71251c765081f7/href">https://medium.com/media/32aeec4821e2582b1e71251c765081f7/href</a></iframe><p><strong>When was the last time you felt an urge to work on something?</strong></p><p>Things change when a group of people challenge the way things currently work. TopDev is doing just that.</p><p>Society doesn’t benefits as much as it could, when you’re toying with that Excell spreadsheet, or working on something that’s outside your area of interest. You do your best work when you actually want to work on something.</p><p>To each, his own, and there’s enough (development work) for everybody. There are enough products to be built, to make everybody rich.</p><p>I won’t talk about how many things go south at your current job. You know it all to well, and i feel no need to bring you down. Not after Bukowski’s poem.</p><p>Let’s talk about the future. Your future, the impact you could be making and <strong>kamiwaza</strong>. TopDev provides work you love doing, and gives you the tools to benefit directly from it.</p><p>How?</p><p>TopDev has put together a <strong>software specification tool</strong>. Each functional component in the product has been captured in a unit or integration test. Also, each functional component has a reward attached to it. Make it pass, and you get the reward. Choose what you want to work on, in terms of mission, impact or technology. What needs to be done has been already laid out.</p><p>Once you get the reward, you can cash out immediately, or you can keep it as a stake in the company.</p><p>Here’s a brief article, introducing the <a href="http://example.com/">technology</a> and another one that serves as a <a href="http://example.com/">demo</a>. Here’s one about the <a href="http://example.com/">economics</a> of the platform. And one studying the effects of implementing this.</p><p>If you’re the entrepreneurial type, you might want to read about how projects get selected, and how you can start your very own.</p><p>Above all, we want to share TopDev with you. The goal is to have the initial co-founders have less than 1% stake in the platform.</p><p>Join us now, do great work and take direct control of your finances, time and challenges!</p><figure><a href="https://medium.com/swlh"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YqDjlKFwScoQYQ62DWEdig.png" /></a></figure><h4>This story is published in <a href="https://medium.com/swlh">The Startup</a>, Medium’s largest entrepreneurship publication followed by +423,678 people.</h4><h4>Subscribe to receive <a href="https://growthsupply.com/the-startup-newsletter/">our top stories here</a>.</h4><figure><a href="https://medium.com/swlh"><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ouK9XR4xuNWtCes-TIUNAw.png" /></a></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=779d9ee66559" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/building-a-playground-for-top-developers-779d9ee66559">Building a playground for top developers</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>