AVC Downtime

For much of yesterday if you came to AVC, you were greeted with this message:

Long time AVC readers have seen this before and it is a sign that something is awry on the shared server that I run WordPress on at Bluehost.

A number of regular readers reached out offering to help me move to a static platform and I will likely take them up on that.

Until then I can only apologize for the availability issues yesterday.

#Weblogs

Comments (Archived):

  1. jason wright

    what would you have posted about today if this hosting problem had not come up? something about the Facebook hearings perhaps, or a more important topic?

    1. Twain Twain

      More important topic? Elon Musk’s co-founded AI startup has this new mission: “highly autonomous systems that outperform humans at most economically valuable work.”https://thenextweb.com/arti…

      1. Lawrence Brass

        The “outperform” part is a bit scary.I hope that by then, every robot will have a senseus © – love and care co-processor installed. 🙂

        1. Twain Twain

          LOL, Lawrence! Alas there’s only 1 of me.Investors don’t care. They invest in block-chaining, automating and efficiency-proofing everything.Not in love and care co-processors.

          1. Lawrence Brass

            Single man or woman shows are hard. More so if you are a thought outlier.As aspirational outliers we must take care of mantaining a functional interface with other people’s realities to avoid losing touch. Deep work take us to places that literally no one has been before or at least no one you know at the same time and place.It is like a scuba diver that one day realizes that she don’t need the diving equipment anymore, and begins feeling comfortable breathing underwater. It is cool but lonely, intellectually speaking. It happens that at those depths the only one who can really make an assesment of your work or run any sort of proof is yourself. In fact it is risky to be so deep alone because the chance of taking damage in the case of an hypothesis failure in the hypotheses’ chain is high.I don’t know if this sounds as bs for most but I guess that crazy people know what I am talking about. And I am sure you know, which implies that I think you are crazy but, you know, the good crazy.So you are in decompression mode relearning to breathe and willing to dive again, sort of? What to do? Reality sucks.

          2. sigmaalgebra

            A lot of that is correct. But for knowing what are doing in some deep dive, there are some solid ways. The ways are applicable only narrowly and, e.g., don’t provide support for wild guesses that get lucky.

          3. Lawrence Brass

            I am very interested in your approach.I used benchmarking for a while to give the process some accountability, but the results kept validating my preferred choices. Totally biased. I was fooling myself.

          4. sigmaalgebra

            Maybe you mean my old work in anomaly detection. Well, it’s a far out statistical hypothesis test: The null hypothesis is that the system being monitored is healthy. When we watch real time data and can reject that null hypothesis, we call that an anomaly and raise an alarm.What is special is that the test is both distribution-free and multi-dimensional. And the work shows a large collection of such tests.There is a simple, intuitive view: In effect a anomaly is a point too far from other points that were collected when it appeared that the system was healthy. So it’s a case of nearest neighbors.So, what’s new, correct, and significant, powerful, valuable? Well, just nearest neighbors is not probabilistic or an hypothesis test. E.g., we have no idea about false alarm rate. So, my applied math was to do a little applied probability manipulation that showed how to know, select, set, and get essentially any false alarm rate might want. So, I ended up with a real statistical hypothesis test. Part of the relevance is that in practice false alarm rate not well known, apparently way too high, and without means of knowing or adjustment is a big problem, e.g., causing people to ignore detections.Well, the applied probability manipulations I did are wildly different from anything I’ve seen anywhere in AI. So, I just did some applied math. There are some pure math prerequisites, e.g., measure preserving as in ergodic theory, some abstract algebra group theory, a use of a classic result of S. Ulam, etc.I suspect that more could be done and that maybe I could clean up much of the math using some stuff from sufficient statistics.But I know with rock solid certainty that there is no real interest in any of the more likely applications — US national security, credit card fraud, monitoring high end server farms and networks, automated stock trading, etc.If I got a lot of venture funding, put together a big, expensive team, wrote a lot of infrastructure software, had a high end marketing effort, looked like a really serious effort. did a lot of alpha testing with some high end sites, blew through maybe $100 million in equity funding, mostly for the infrastructure software and the marketing, then, maybe, I could construct a company worth $1 billion. Maybe.I did the work and published it. The math is correct. The work totally blew away the work on monitoring we were doing in our AI project. So, I’ve done my part. Other people, no matter how much they are concerned about early detection of anomalies, high detection rates, low false alarm rates, just are not much interested.With a big publicity push, it’s possible to sell empty bottles of air, pet rocks, ice to Eskimos, and even AI.But I have a much easier way, also from some applied math, to make much more money much more easily, cheaply, etc. If my startup works, then, sure, I’ll deploy my anomaly detectors all over the site. But for now, my anomaly detection work is hidden in plain site and totally ignored — I’ll leave it that way.

          5. Twain Twain

            Thanks. Yes, I’m an outlier. DeepMind does the most cutting edge AI, right?Its Head of Machine Learning tweeted about “crass binary classifiers” in Feb 10 2018 and I’ve already solved that problem.https://uploads.disquscdn.c…And now with FB’s problems …https://qz.com/1250605/dont…The industry hasn’t been able to solve it …* https://www.fastcompany.com…The market NEEDS my system.Meanwhile, investors thought the opportunity was in blockchain. Nine years and $billions later …https://uploads.disquscdn.c…Now, if they’d given a $1 million to me in 2010, some of the problems related to data structures and AI would have been solved by now.I’m not the one lost in the deep ocean. The techcos hit the iceberg and I’m in my little spaceship flying past.LOL!

      2. sigmaalgebra

        Totally hopeless goal. Taking some 50+ year old regression analysis in statistics and calling it “intelligence” or “learning” is just hype. Even if every newsie in the world runs the sensationalist headlines, the goal is still no closer. Musk’s statement is just 100% hype. Doesn’t the SEC have some rules about such things?

        1. JamesHRH

          Your shortest comments are your best and yet, as is true with all of us, they might be more powerful if they were even shorter:Humble suggestion:regression analysis /= intelligence or learning.

          1. sigmaalgebra

            The issue is more than just the inequality. Instead there is the goal which, IMHO, is hype, e.g., deceptive. So, again, maybe the SEC should get involved. Then a big question is, how come the hype gets so much attention? Well, “In the morning a lie has flown half way around the world while the truth is still getting out of bed.” That is, the newsies want sensationalism.So, as we all know very much too well, (1) maybe some people know that some hype is coming and take a highly leveraged long position in some relevant stock, (2) the self-serving hype is announced, (3) the newsies spread the sensationalist hype far and wide thus “pumping” up the stock, (4) the longs close out, dump”, their position. As you know, it’s called “pump and dump”. It’s also “buy on the rumor and sell on the news.”. So, maybe there are traders sitting at Bloomberg terminals waiting for a new Musk statement then go long, wait two days, dump and go short, wait two days, and close out and wait again.And now we have Zuck telling some gullible Members of Congress that some of their concerns will be solved with some “AI” software. So, some of the hype is taken for real, as the basis for a promise, as hope, and maybe to forestall legislation.You are right: We’ve known that the newsies were rotten all the way back to Jefferson’s bitter denunciation. So, I should just accept that the hype, sensationalism, newsie junk, pump and dump, etc. are just reality, avoid them, and f’get about them.

        2. Lawrence Brass

          The new is the high volume and freshness of the data being analyzed and the low latency of the inputs and probably the outputs.Systems now can react on data in lively, scary ways, even enter in resonance as it has happened in automated trading systems.That makes a big difference from reading a statistical report from last year.Isn’t that low latency and reactive behavior a part of our own so called intelligence?

          1. sigmaalgebra

            All earth based intelligent animals breathe. If they don’t breathe, then they cease to be intelligent. So, in this sense, breathing is “part of” intelligence. But all animals breathe, including the zooplankton. Okay, okay, I’ll agree: The zooplankton are as “intelligent” as AI!!!!> Systems now can react on data in lively, scary ways, ….That’s been true in systems of wide variety — in US national security, process plant monitoring, computer system monitoring, just the monitoring in passenger cars — for decades, but those systems are not in any sense “intelligent”. E.g., there is (A) real time monitoring and anomaly detection, (B) systems to responding automatically to such detections, (C) real time deterministic and stochastic optimal control, etc., and I’ve contributed to all of (A)-(C), but there was nothing “intelligent” there.Supposedly there is now a wild example: Rolls-Royce got into high end aircraft engines early in WWII, e.g., the Merlin V-12. Then they continued with jets. Now they have high end high bypass turbofan engines. So, if have a plane, say, just to pick a number, a Boeing 757 with two of those Rolls-Royce engines, supposedly the engines, at least any time they are running, are in constant real time communications with monitoring computers back at RR’s HQ in England. Amazing? Yes. Intelligent? No.

          2. Lawrence Brass

            Very strict reasoning Dr. Sigma.We carry a lot of interconnected bioelectric, biochemical, autonomous, analog-digital.. systems in our bodies. Our current models of intelligence are still very crude and limited.I love Rolls-Royce as a company and the Merlin. I’ve written here before that my father assembled and installed Merlins at a secret RAF base in the UK during WW2. So he told us. According to him they would work until they fell asleep and then the others in the shift would pull them over and continue working, non stop.Single crystal turbine blades.. the result of pure glorious and amazing interconnected intelligence, craft, and engineering.https://www.theengineer.co….

          3. sigmaalgebra

            Yes, I saw a long movie documentary on some of what RR was doing with their engines, including the turbine blades. Amazing.At times the English and their stiff upper classes look like a bunch of total, hopeless, losers. At other times they are way up front with some of the best work in the world.Yes, the Merlin, the Spitfire, and the British work in the Battle of Britain were amazing: From the outside, and even the documentaries, it appeared that just somehow that plane and engine got designed, both world class, and produced with high quality in huge numbers right away.But for that success, I can believe the approach to sleep: Work. Work until hungry and/or thirsty, grab something to eat and drink but keep working, work until flop, sleep a little, and work again. Apparently that and more is what the British did.At times I’ve done things like that: Work, order pizza delivered and keep working, clench teeth hard enough to crack molars, flop down and get some sleep on the floor under the desk, wake up at dawn and continue. Shave while working. F’get about changing clothes! I discovered that working like that can lose weight no matter how much pizza, cinnamon rolls, etc. eat!That current high end RR engine likely involved a lot of working at the plant 24 x 7 with a few hours of sleep each night. RR and the British deserve high praise.

        3. Twain Twain

          It’s not Musk’s statement. It’s OpenAI’s mission statement. Here’s where their AI is. It’s more than regression analysis:* https://www.youtube.com/wat

      3. JamesHRH

        Good luck.

        1. Twain Twain

          Thanks, James.OpenAI has a $1 billion to achieve their mission of building highly autonomous AI that outperforms humans in the most economically valuable work.Unless another group of investors invests $1 billion to build AI that’s about augmenting rather than replacing/displacing humans, OpenAI’s model will be the only game in AI.

      4. jason wright

        AI sounds a lot like a forming religion.

    2. LE

      You have to love the fact that someone who is worth over $60 billion dollars actually has to work that hard (prepping and getting ready) and be under the microscope like that.[1] If having money doesn’t buy you the ability to avoid discomfort, what does it buy exactly? The same guy who pitched investors in a hoodie when he did the road show.[1] Down to the booster seat. Missing the next day. He should have kept it and made a joke about it when he started on the 2nd day. Hold it up in his hand. YES I BROUGHT THE BOOSTER SEAT AGAIN! Everyone would laugh. And that would be the lede.

      1. Lawrence Brass

        I was expecting he would perform better. The venue, being seated there has to be tough.Anyway, 100 X more honest than James Clapper in the same seat.It is time to bring Snowden back, give him a chance.

        1. Salt Shaker

          Only saw the 1st day where Zuck performed well, in part cause the folks asking the Q’s didn’t seem particularly well informed about his biz. Second day, so I’ve read, was a little tougher for him. This shit isn’t going away as long as folks out there can exploit personal data, regardless of what regs come about, though I think the U.S. gov’t (not Euro) will give fb some rope to figure it out on their own. The optimal circuit breaker would require a clear, unobtrusive opt-in button (rather than opt-out), which would seriously damage fb’s (and others) biz model. Conversely, the quality of an audience that does opt-in, if such a button were to exist, could be of more value and generate higher cpm’s, though at the expense of scale. No question fb’s gonna also explore a sub model for those who want it, but to really succeed they’d need to enhance their value prop. All that aside, it does seem like a runaway train w/ respect to personal data, with credit services imo on the forefront.

          1. LE

            They will do things that will allow them to say they are doing things and appear to be doing things so that people checking up on them can check all the boxes [x] and appear to have made them do something. ‘We did our part’. In the end ‘normals’ don’t care about this. It’s something that people who know about and think Kara Swisher is great pretend to think is a big issue. It’s driven by the media as is all evils.It’s funny. You put all this personal shit on Facebook or you write a blog or are on twitter and then all the sudden you care that some info is shared about you or bought whatever.Here let me give the play book on this. This is what Matt Blumberg didn’t do unfortunately (some of it I mean he did ‘a’).a) Pretend to show genuine concern and deference and accept blame in a ‘buck stops here’ fashion.b) Act like you are working on a solution. Work on a solution.c) Buy time but show some progress doll it outd) Be willing to update and answer questions and so on. Be transparent.e) Expect that in short order people will no longer care all the steam will have vented.Problem solved.

        2. JLM

          .Clapper, Brennan, Comey, McCabe, Rosenstein, Mueller, Lynch, Holder belong in jail.JLMwww.themusingsofthebigredca…

      2. PhilipSugar

        The biggest difference in working for a big company (Government is the biggest) is that you really do have to prep and get ready.At first I thought what a waste. But you really do have a staff to just do that because as you say you will be under the microscope.

      3. JamesHRH

        Its not the money that has him on the booster seat, its the philosophy of misdirection and lack of transparency.Don’t see Tim Cook there, FWIW.Oh yeah, and the poor performance worked perfectly. Looking bad by saying his team would follow up is no big deal, as his team will not follow up and they will go back to scraping everything hey can about users and most of us won’t care at all.

        1. LE

          Cook is an idiot. How can someone, with a straight face, tell people that he could make money in that way (selling data) but choose not to ‘pat me on the back’. The fact is Apple makes money by way of selling extremely profitable hardware so it doesn’t need to. But make it sound like you are just so above it all. Whereas FB is a free product that sells ads.

          1. PhilipSugar

            “Pride goeth before a fall”I’d be damn careful making statements like that.Here are some gems from their Privacy Agreement:Apple shares personal information with companies who provide services such as information processing, extending credit, fulfilling customer orders, delivering products to you, managing and enhancing customer data, providing customer service, assessing your interest in our products and services, and conducting customer research or satisfaction surveys. These companies are obligated to protect your information and may be located wherever Apple operates.We may also use personal information for internal purposes such as auditing, data analysis, and research to improve Apple’s products, services, and customer communications.Apple’s websites, online services, interactive applications, email messages, and advertisements may use “cookies” and other technologies such as pixel tags and web beacons. These technologies help us better understand user behavior, tell us which parts of our websites people have visited, and facilitate and measure the effectiveness of advertisements and web searches. We treat information collected by cookies and other technologies as non‑personal information. However, to the extent that Internet Protocol (IP) addresses or similar identifiers are considered personal information by local law, we also treat these identifiers as personal information.We may collect information such as occupation, language, zip code, area code, unique device identifier, referrer URL, location, and the time zone where an Apple product is used so that we can better understand customer behavior and improve our products, services, and advertising.We may collect information regarding customer activities on our website, iCloud services, our iTunes Store, App Store, Mac App Store, App Store for Apple TV and iBooks Stores and from our other products and services. This information is aggregated and used to help us provide more useful information to our customers and to understand which parts of our website, products, and services are of most interest. Aggregated data is considered non‑personal information for the purposes of this Privacy Policy.We may collect and store details of how you use our services, including search queries. This information may be used to improve the relevancy of results provided by our services. Except in limited instances to ensure quality of our services over the Internet, such information will not be associated with your IP address.With your explicit consent, we may collect data about how you use your device and applications in order to help app developers improve their apps

          2. LE

            I might make enemies for expressing an opinion here.My theory is that in addition to all of the valid reasons why Tim Cook cares about privacy there is another one that nobody talks about. It is because Tim Cook only ‘came out’ in 2014. So here we have a man (in an older age group raised in fear) who grew up in the closet. As a result he most certainly had to be more sensitive to privacy than the average citizen. It would only be human nature.Another thing that I have found over time is that people who break the law in some small way (say they smoke pot) will also be paranoid in the same way about privacy. How could that not be possible? Ditto for file sharing. And note that some of these issues are more likely in the tech crowd than with normals.Remember when you first learned to drive how hyper alert you were about cops and getting a ticket? Because if you did your parents would bust your ass? It’s sort of the same thing.

          3. scottythebody

            Why is that “idiot” level? I guess you could argue hardware is a risky business given the future you see, or that Apple sucks at online and Siri (Which they do) and probably could do better if they traded in their customers’ privacy, but it’s also really smart to *not* be in the advertising based on customer data business. The biggies are, what, Microsoft, Apple, Facebook, Amazon, Google… Forty percent of those are advertising companies. So I see it as a strong differentiator. Apple is the only hardware company.

      4. JLM

        .They never layed a glove on Zuck. He made them look tech illiterate.Real men don’t use booster seats after age 2 or when they stop breast feeding.Advantage Zuck.What we witnessed was the postmortem of the lynching of privacy.Zuck contributed to almost every Dem who pretended to quiz him.Only Teddy the Cruzer touched Zuck.JLMwww.themusingsofthebigredca…

    3. fredwilson

      i didn’t watch the hearings and don’t really care. that was obviously going to happen at some point. i think blockchain/crypto/decentralized internet will help address it in time

  2. JLM

    .This is why I come to AVC.com.JLMwww.themusingsofthebigredca…

      1. Lawrence Brass

        To fill in, as anchors do when the satellite link goes down.@JLM, anything to say about the current non-events?

        1. JLM

          .Syria going to be very dicey. US and Russian planes will tangle. US kills Russians.Russians will attack a US cruise missile launch vehicle.Navy will splash every Russki plane. Go Navy!Mueller and Stormy elope to Huatulco. Turns out Stormy is trans.Comey calls Pres Trump a “Mafia Don” and The Don says, “Comey, that stiff, finally got something right.”Turns out the EC was written by Strzok and is nothing, humiliating the FBI/DOJ et al. Rosenstein shuts Mueller down.Pres Trump calls out NY Nat’l Guard to protect his doctor’s and accountant’s offices.JLMwww.themusingsofthebigredca…

          1. Lawrence Brass

            By the current standards, looks like a normal day.Thanks for the report! 😉

          2. Michael Elling

            Did one consider that current Syria crisis is orchestrated to deflect attention? You’ve got quite a director there!

          3. JLM

            .Of course, because Pres Trump is a rotten SOB, right?I doubt even Pres Trump has the gear to gas some innocents. Who knows?I am always amazed that the Russians always plead that the gas attacks either never happened or were perpetrated by others. If others, they would be the people we are supporting who don’t generally operate in that region of Syria.The Russians could have bought their way out of the dilemma by approving the UN Sec Council plan to inspect, but they vetoed it.The shelf life of binary gases (different than “persistence”) is notoriously low because attacking troops need to enter the areas they have gassed, so the presence of a formulation plant is a key to identifying the bad guys.In the last gassing, it was established that the Syrians had a mixing plant at the airport they used.Donald J Trump owns the news cycle, the space between the MSM’s ears, and the messaging. That is not a new development.But, no, I don’t think Gen Mattis and Gen Kelly and Dir CI Pompeo are in on a fix.JLMwww.themusingsofthebigredca…

  3. Erin

    Mercury retrograde. 🙂

    1. fredwilson

      exactly.

    2. mplsvbhvr

      I can’t take much more….

  4. Susan Rubinsky

    Classic WordPress failure. Maybe time to switch to a more robust CMS. Drupal would be my choice.

    1. fredwilson

      def not doing that

      1. Gregory Magarshak

        Qbix Platform that I built is way better than Drupal etc. It solves these kinds of issues.SAFE network will be even better!!! If it ever launches. For now there is only Freenet and it’s been around for ages, and not very mainstream.Maybe you can ask your friends at IPFS to put something like SAFE network out there, hosting blogs is popular (20% of new Web sites are powered by WordPress) but static blogs running on an automatically scalable uncesorable permissionless system is a killer app.I wish you could invest in SAFE network but they are too close to IPFS.

  5. DJL

    Wait a minute! Your portfolio company Cloudflare should stop this from happening. They cut their teeth protecting WordPress.

    1. LE

      Yes exactly further the content is static and the comments are delivered by disqus. But this assumes everything is configured correctly.https://support.cloudflare….Part of the issue is making sure CF doesn’t cache things in a way that makes you wait for the latest post to appear. In the past this has happened with AVC. In order to get the latest post you had to arrive by Fred’s twitter link. A few times (and there could have been other issues under the hood) you might not see it for several hours.

    2. fredwilson

      i wish Cloudflare would just offer hosted WP and then i would not have to figure out who is causing this issue.

      1. DJL

        Agreed. I had to go through the SAME thing and actually punted Cloudflare to save my site.

  6. ShanaC

    Oh, glad it isn’t just me seeing this

  7. Kirsten Lambertsen

    I strongly recommend Pressable. They’ll do the migration for you. Feel free to ping me if I can help. As you know, I know a little about WP 😉

    1. fredwilson

      others have told me to migrate to WP Engine. what do you think of that advice?

      1. Kirsten Lambertsen

        It’s probably fine, but I do think Pressable is better. Pressable is owned now by Automattic, which means they have to meet the highest possible security and performance standards. Even though I’m bound to be biased, I also genuinely believe the customer support is far superior.More than happy to answer any other questions or help however I can 🙂

  8. Miriam Schwab

    Hi Fred! Our solution, Strattic, can solve this issue for you now, and prevent it going forward. We publish your WordPress site as an exact static replica of itself in one click. This makes the site faster, scalable, and super secure. https://www.strattic.com/

    1. fredwilson

      do i migrate from Blue Host or do you mirror it somehow?

      1. Miriam Schwab

        We create a clone of your WP site on our secure staging area, and then deploy a static version that’s an exact replica and behaves exactly the same, just no database to cause issues, slow things down, or be a point of attack. That static version is served up via CDN for even more speed and scalability (plus http/2 and gzip, and https without mixed content – you currently have mixed content on your site). You point an avc.com CNAME at our CDN and from that point on people visit the static version of your site, which provides the exact same experience as your “real” origin site (except faster and more secure). When you want to write new posts, you login to the secure staging version of your WP site, write your post as usual, and then when everything is ready you click the Strattify button and it gets deployed to the static site for all to see.Basically you have three instances of your site running at all times:1. Real WP site (only accessible to you) >> 2. Static replica >> 3. CDN which is your publicly facing site.This means your site is super stable and scalable.We can set this all up for you without affecting your current site or taking it down so you can see how it works. If you like it, you point your CNAME. If not, no problem because there’s nothing to undo.

  9. Vasudev Ram

    Static site generators (SSGs) are great in many ways, though of course they have some cons too. They consume much less machine resources, and have other benefits, such as less potential security issues vs. dynamic blogging software such WordPress and others. (The dynamic ones can have many more features though.) Threads about SSGs come up periodically on HN. There was one recently:Show HN: Makesite – A static site generator in 125 lines of Python (github.com)https://news.ycombinator.co…I mentioned my own SSG called PySiteCreator (experimental version, not production-ready) here:https://news.ycombinator.co…in that thread.Here is a blog post about it:Early release of PySiteCreator – v0.1https://jugad2.blogspot.in/…The key benefit of the approach I took is mentioned in these two paragraphs from the post:[ A useful feature of PySiteCreator is that it allows you to write almost any Python code you wish, that can fetch data from any source, for outputting to the created HTML files. The data can be fetched from text files, CSV files, DBM files, any type of file in the file system from which you can fetch data by any means, from a relational database using the Python DB API, from any source on the Internet, using sockets, XML-RPC calls, HTTP REST calls, SOAP calls, CORBA, PYRO, RPyC, or any other suitable method.There are only a few conventions to be followed, such as importing the psc_util module in each Python file you write, and defining a top-level function in each file, that should be named create(), in order to be able to use PySiteCreator. Other than those conventions, there are no other restrictions or requirements about what kind of code you write in the Python files. ]Searching for “static site generator” in the search box at the bottom of any HN page will find many such threads about them.SSGs where you have to do some manual work to set them up and use them, probably won’t suit Fred who may be looking for something more plug-and-play. But they can be great for people who can invest a bit of time to do that setup and usage.SSGs also play well with sites with free hosting like Github Pages. Just build your site locally with usually a single command (each time you add or update the content), and then push to Github Pages with a regular git push command, same as what you use for software projects. Or you can upload the content via the Github UI too.

  10. Matt A. Myers

    Bluehost is owned by Endurance International Group – https://www.reviewhell.com/…They have a scam setup to upsell SiteLock.