Mark Gritter's Journal|
[Most Recent Entries]
Below are 20 journal entries, after skipping by the 20 most recent ones recorded in
Mark Gritter's LiveJournal:
[ << Previous 20 -- Next 20 >> ]
[ << Previous 20 -- Next 20 >> ]
|Monday, February 11th, 2013|
|Five Things Make a Post
I am somewhat willing to believe the Advertising Federation of Minnesota, when they say that advertising taxes don't work.
But the arguments for "why tax X is bad" seem cookie-cutter. Drives X across the border? Check. Money that is taxed can't be spent elsewhere in the economy? Check. X is already taxed through personal income and sales taxes? Check. I'd like these arguments a lot better if we admitted that, yes, taxes are bad, but we are looking for the least-bad solution. Then it might not come off as special pleading. (Or non-special pleading since the arguments seem always the same.)Two. Qonqr gets a mention in the Minneapolis/St. Paul Business Journal.
Shorter version: once Scott gave up finding capital and just worked on building the game, he was able to get more traction building the game! Also their yearly revenue is about $150k, which is decent for an indie freemium game but still pretty poor for a company listing six founders.Three.
VMware continues its acquisitions centered around the software-defined data center, announcing the purchase of Virsto (for an undisclosed amount.) Virsto makes a virtual storage appliance that pools conventional storage arrays into a smarter pool of storage for VMware (and, for the time being, Hyper-V.) Their file system has some nice VM-aware features which might influence the direction of VMware's Virtual Volumes (vVols) technology. News reports selling this as a "growth" are bogus; this is a play to broaden VMware's infrastructure story but it will not sell any additional VMware installs in the short run. VMware has already dipped their toe into this sort of technology with their vSphere Storage Appliance (which pools local disks into VMware storage.)
Virsto has $24m in venture backing providing a likely floor for the purchase price. (Virsto has somewhat of a strange corporate history with their $12m series B round partially going to make an acquisition of Evostor....) Their offering is software-only with one reference quoting a list price of about $2500/host. No real lead on revenue numbers, but I wouldn't expect to see a big multiple here. (It's hard to get big deals when you're not selling the disks--- but, your margins are good.) VMware's license list price is between $1000-$3500/host, but the combination of the two products will not result in an additive increase.
The model of "smart software aggregating dumb storage" has a certain appeal, but I think you will see some roadbumps-- that would make a whole other blog post. (Short version: even dumb storage isn't dumb these days, but the storage APIs do not provide sufficient visibility to control them.)Four.
I read J.J. Adams' new anthology of mad scientist stories, reviewed by mrissa previously
. Some of the stories were just dumb (Harry Turtledove obviously doesn't know what makes a funny story actually funny) but there were enough good ones to keep me entertained. And yet... most of the authors don't seem to come from a background very familiar with the practice
of science and engineering. Where are the mad scientists' grad students? Or their patent lawyers? Mad venture capitalists hanging around? Mad scientists probably need to be administrators as well as innovators to get any proper mad science done these days!Five.
My insurance company actually cut me a check! For overpaying last year when Tintri switched insurers after we'd met the family deductible. But not the complete check, since the prescription handler (ExpressScripts) seems to be more efficient than Anthem itself at getting things done.
|Sunday, January 27th, 2013|
I finished James Treadwell's "Advent" which is a fantasy novel that really tries to emphasize how creepy and horrific modern people would find the emergence of magic. So it's almost horror-tinged. Unfortunately what this did for me was make all the characters unsympathetic wibblers. It's a decent story, though, and I'm interested enough to see where he's going in the sequel.
Unfortunately the mood was really quite ruined by a big infodump 2/3rds of the way through the book... mainly of things that were already implicit (I felt) or didn't need to be explained anyway. Also his 16th-century Faust (German of course) seems to carry on conversations with his spirits in modern English--- mostly this was OK (I could just read it as translation) until a third party could overhear him.
|Friday, January 18th, 2013|
In an arithmetic sequence --- for example, 1, 4, 7, 10, 13, 16, 19 ... --- each number is the arithmetic mean of its neighbors. (In fact, it is the arithmetic mean of all of its N predecessors and N successors.)
In a geometric sequence (1, 4, 16, 64, 256, ...) each number is the geometric mean of its neighbors. And again this can be expanded to the N closest neighbors.
What other "aggregate" functions can produce sequences of this sort? For example, "max()" and "min()" produce only the constant sequences.
The "mex" function (minimum excluded number) produces a cyclic sequence, but cannot be expanded from one neighbor to multiple neighbors. For example,
mex( 0 ) = 1
mex( 0, 1 ) = 2
mex( 1, 2 ) = 0
produces the sequence 0, 1, 2, 0, 1, 2, .... Taking two neighbors produces the same sequence, a promising start!
mex( 0, 1, 0, 1 ) = 2
mex( 1, 2, 1, 2 ) = 0
mex( 2, 0, 2, 0 ) = 1
But mex( 0, 1, 2, 1, 2, 0 ) = 3. In fact I think we could show that 0, 1, 2 repeating and 0, 2, 1 repeating are the only possible mex-based sequences.
How about 'median'? Or '90th percentile'? Median works fine for any increasing sequence, if you include the number itself.
In general, what conditions on F: 2N
-> N are necessary and sufficient so that a sequence s0, s1, s2, s3, ... exists with F( sn-k
, ..., sn
, ..., sn+k
) = sn
? What if we require the sequence to be nonrepeating?
|Sunday, January 13th, 2013|
|Dear advantage players,
What is the best way to get some value out of my crappy Citicards "easy deals" points?
The rewards site offers a variety of gift cards and merchandise for 10-30% discounts, with the points translating dollar for dollar to the discount amount. Buying dozens of $50 gift cards for $40 each and reselling them seems not worth my time. Are there any options for bulk resale?
The only points-only items are available through auctions, and the auctions seem calculated to be as annoying as possible. Your only option is a fixed-increment bid increase, so the winning strategy seems to be to wait until the last minute and click like crazy. (Oh, and the conversion rate is generally 10x worse.)
Long-term solution: switch to a new credit card with a decent rewards program.
|Tuesday, January 1st, 2013|
|Storage won't be commodity
VentureBeat has some 2013 predictions, including one from Bipul Sinha of Lightspeed Venture Partners
. (Lightspeed invests in Tintri and several other storage startups.)
“The enterprise storage market is experiencing a tectonic shift led primarily by virtualization and storage media disruptions. The incumbents are slow to respond, and the startup activity is at an all-time high. The new year will witness two significant trends: an acceleration of storage and compute hyper-convergence and a bifurcation of performance and capacity tiers for shared enterprise storage.
“These trends will eventually commoditize the storage layer of the enterprise information technology stack and essentially propel the emergence of software defined, agile enterprise datacenter.”
It doesn't seem to me that both
these points can be true. If the storage layer is really going to become commodity, then VCs are just burning money by backing anybody in the sector. On the other hand, if virtualization requires a different set of storage assumptions, behaviors, and controls--- or if the future is converged compute and storage--- then the resulting products will hardly be interchangable "commodity" components!
I do think Bipul is correct to point out the separation occurring in the storage market, but it is into three tiers, not two. High-capacity "bulk" storage is one part of the market, and ultra-high performance storage operates at a quite different point. But the majority of enterprise workloads don't fit neatly into either of these categories. The software-defined data center will require storage that is virtual-machine aware
, operates well over a broad range of applications, and requires little in the way of customization or management. This application-centric storage tier is where Tintri focuses.
|Sunday, December 30th, 2012|
|Best Buy in a negative light
This Star Tribune coverage of Best Buy's failed buyback program and $27m award to TechForward
paints a very negative picture of Best Buy, but seems to come puzzlingly late. The case concluded December 5th, and this article doesn't seem to add any new information. Best Buy has not (yet) filed an appeal, so there's no progress on the legal front, just some high-level commentary by a local law professor. Nor is there any additional information about the buyback promotion's fate.
Although one shouldn't read too much into court arguments as statement of company policy, there is one particularly poisonous rationale that I hope Best Buy doesn't try to include in its appeal:
"The jury has made TechForward a highly profitable entity, a result that TechForward was never able to achieve in the marketplace," Best Buy said in a court filing. "TechForward never turned a profit and sustained itself through the largess of its investors."
See, VC-backed companies aren't "real" like Best Buy, so stealing from them shouldn't be punished as much?! A startup company like TechForward is often building towards profitability rather than profitable from the start, and signing a big partner like Best Buy can make all the difference. So for BBY to claim that the company was a failure anyway--- after BBY broke agreements and stole their IP--- sends a clear message to any future startup partners. If that's the attitude of top corporate officials, then it's no surprise that lower-level employees won't respect their partners either, whether or not "authorization" is given to engage in theft. Best Buy should realize that the monetary award is not the only thing at stake here. Many local entrepreneurs and potential future partners are watching as well.
|Wednesday, December 26th, 2012|
I was craving some tortas (as I didn't get any in California last week) so Marissa and I went up to Andale Taqueria in Richfield. The tortas were satisfying, and the bread was good, although perhaps they were a bit stingy with avocado. Of course, comparing to California in this regard might not be entirely fair. Marissa had carnitas with adobo sauce, and I had carnitas al pastor. I was somewhat surprised to get finely cut pork instead of thin slices, but the flavor was quite nice. Worth a return visit if the craving returns, or I want to introduce somebody to Mexican tortas. (Mexican tortas are served on an oblong loaf of bread, usually with onions, lettuce, tomato, cheese, peppers, and mayo. Other countries' tortas may be omelettes or other concoctions.) Also they have tamarind Jarritos, which is the best kind of Jarritos.
My research also found Manny's Tortas in the Midtown Market. Numerous food blogs praised "Tacolmeca and Supermercado Olmeca" in Burnsville, which is now departed. There appears now to be a (re-opened?) Olmeca Market on Irving Avenue in Burnsville, but I don't know whether they have prepared food or are just a grocery.
|My insurance company decided I was too low on stress
Anthem Blue Cross has already made it perfectly clear that they don't pay any attention to messages in their electronic message center. After a request sat there for a week, I emailed them. Three days later they answered a month-old query instead. They also have been dragging their feet on crediting my deductible for the first part of the year (Tintri switched insurers in October.) But, hey, those are both par for the course for an insurance company.
Today I got a "Claims Information Request" form, in a hand-addressed envelope, which had none of the items checked or any information filled in except my name, address, and member ID. It looks like something they should have sent a provider instead. I called (having learned my lesson above about electronic communication) and, as expected, the call center worker had no idea what the heck this was, and said that there weren't any rejected claims.
I am puzzled at what process could have resulted in this behavior--- perhaps the person who should fill out the form is not the same person that addresses them and puts them in the mail?
|Thursday, December 13th, 2012|
|Video interview by Kieran
Kieran Harty, Tintri's CEO
, talks about coming up with the idea, starting the company, finding a co-founder (me!) and bringing the product to market.
He mentions that we met outside at Sun. It was a nice day, but I screwed up in two ways--- I didn't reserve a conference room or visitor badge (I'd never tried to do either at the Sun campus) and then the outside place I thought had seating turned out to be rather uncomfortable. Fortunately Kieran forgave both this and the later ketchup-splatter incident. :)
|Tuesday, December 11th, 2012|
I acquired Triple Town
recently. The PC version has both the game itself and a "metagame" of collecting items and building up your capitol city. The base game itself is a pretty fun puzzle, but the reason I looked at it was because of the marketing (on Steam) as a play-and-build game. The building is pretty limited, though.Puzzle Pirates
is the iconic version of this for me. There's individual puzzles for low-level tasks. But these fit together to provide fodder for the higher-level games of trade, shopkeeping, sea battles, flotillas, and even dress-up. It succeeded at making the minigames both interesting and "matter".
Plenty of games have a PvE grind follwed by PvP endgame (or PvE resource-gathering for PvP battles) but these don't really feel like the same sort of thing. KadoKado
offers access to new games in exchange for coins earned in other games, but I wouldn't call that a "game" at the upper level. In college we joked about having a game where you could play at the SimCity
, the Civ
, or the Master of Orion
level of the same universe, with benefits flowing in both directions--- so Spore
offers something like that but the levels are spread across time. Deck-construction games come close, particularly when success in the lower game expands your deck.
What other examples of multi-level games can you provide? (And are any of them worth playing?)
Two of tonight's Minnedemo presentations were home-automation startups. Spark Devices makes a lightbulb socket (a shim!) that is wirelessly controllable, and Smart Things is building a whole platform for controlling such things.
The idea of home automation has been for a long time, but I'm not clear what the killer app for it is. I know that home security services are moving in this direction a little bit in terms of turning on the lights for you, and I can see some benefit there. And one of the founders talked about his deaf father needing signals about the phone ringing or other noises that could be translated into light toggles. But both those are pretty crude examples.
The thing that seems "neat" to me, but perhaps not compelling enough, is to improve energy efficiency by controlling lights and heating based on your current location. A home automation system could lower the heat when everybody is out of the house, but increase it back up the closer you got to home. Following you from room to room and switching lights would also have some benefit (but requires much more fine-grain sensors--- can't just piggyback off your smart phone.)
Further-out ideas like fridge inventories (requiring RFID tags on all your food?) seem less appealing to me, and signals on door opening/closing can already be adequately handled by dedicated security systems. There's a "do it yourself" market, of course, but it is small.
|Tuesday, November 27th, 2012|
|Lottery still -EV
Powerball is estimated at $500m (annuity) or $327m (lump sum) for Wednesday's drawing.
The jackpot odds are 1 in 175,223,510, so the EV of the lump sum payment is $1.87. The smaller prizes bump that to about $2.21, which looks good on paper for a $2 ticket. But the ~42% state and federal income taxes in Minnesota bring the net value down to about $1.28, even before estimating the probability of a split prize.
Unless, of course, you could put your lottery tickets into your IRA. :)
|Monday, November 26th, 2012|
|App market numbers don't add up
The STrib ran this (national) piece on the mobile app industry: Mobile app industry takes flight
. One of the claims is that mobile apps have been responsible for about 500,000 jobs. This seemed high to me. The article also claimed mobile app developers were earning $90-130K next year. Even $100K average would mean $50B in payroll, which seems too high.
I don't know a source for good public numbers on the app market, as most reports are very forward-looking. $25B for 2012
seems reasonable. However, Apple, who is a clear market leader, pulled in about $1.9 billion for all iTunes sales in the 3rd quarter
, which is compatible with a $25B market. (Their take is 30% and they have about 70% of the app market, but apps aren't all of iTunes.)
There are various possibilities. It could be that most app developers (particularly independents) are satisfied with a far smaller income. Or, it could be that app developers are losing money as a whole (perhaps funded by VC money in anticipation of future growth.) Or, the 500K jobs number is fundamentally bogus, either inaccurate or double-counting.
For comparison, Zynga has ~3000 employees at about ~$300m quarterly revenue. $400K revenue/employee is in line with other tech companies. But if the industry is as successful as Zynga, on average, then that would be less than 100K jobs.
|Sunday, November 25th, 2012|
|What is up with this story?
The news has been circulating that the Obama administration scrambled, before the election, to put the U.S.'s assassination program (oops, sorry, drone strikes) on a more formal footing
, and "resolve internal uncertainty and disagreement about exactly when lethal action is justified."
There are just so many things wrong here. Why is the administration leaking this information, when it won't admit there is any drone strike program in place in any official capacity? (Selective leaking like this makes a mockery of FOIA, not to mention journalism.)
Given that Obama didn't feel the need for any formal legal framework before ordering attacks, what makes him think that Romney would? Further, what makes him think that the Romney administration would care
what guidelines the outgoing officials had laid down?
And, given the scope and duration of drone strikes, what the heck took them so long to come up with legal fig leaves? I mean, the Bush administration's legal reasoning may have been crap, but at least they could turn it out on a reasonable schedule.
I'm trying to come up with a motivation here. I don't seriously believe that the administration is trying to signal it wants more cover from Congress, but it's a possibility. The other potential stance is that somebody believes drone strikes are so important
that any interruption during a handover of power (due to questions about who authorizes strikes and under what circumstances) would be dangerous. Neither of those alternatives makes a whole lot of sense, though.
Best typo from "The Signal and the Noise" so far:
"At NASA, I finally realized that the definition of rocket science is using relatively simple psychics to solve complex problems."
|Thursday, November 15th, 2012|
|David Braben at it again
A Kickstarter project to produce "Elite:Dangerous".
I wish I was more excited about this.
Part of the problem was that "Elite: Frontier" didn't really recapture the magic of the original. The "more-realistic" space combat was just less fun.
However, the other thing that bothers me is that Braben doesn't seem to realize he's competing with a whole generation of people who played Elite, loved it, and already
decided to go write some big open-world space opera. Like, say, EVE Online. Or Notch's new project
. Or X3? Or Spore, if you want to go the procedurally-generated content route.
There's certainly room to do more in this area. Their ill-formed "gee, wouldn't it be neat" presentation of what he views multiplayer would be like doesn't give me a lot of confidence, though.
ETA: I decided to back both this project and Chris Roberts' Star Citizen
despite the very long lead time in both cases.
|Sunday, November 11th, 2012|
|Politico doesn't get an answer
Alexander Burns tries to explain why Republicans' polling numbers were so far off.
The story he tells is that their likely voter models underestimated Democratic turnout.
That's a reasonable answer for why there's error, but the question is not "why were your polls wrong", the question is "why did you trust your model despite overwhelming evidence that few public polls, nor most poll aggregators, nor the betting public, were making the same assumptions?" That's
the real dirt that I hope comes out in campaign books. Was somebody deliberately releasing only the most optimistic projections? Or was there just a bubble that prevented those assumptions from being questioned?
(FWIW, Nate Silver analyzed the last three weeks of polling and found a R-ward bias in almost all of them
, which may suggest Obama's numbers were improving in the last week or so of the campaign. But he says the bias is "probably more than can be explained by the late shift alone.")
|Lexington may be in a state of denial, too
The Economist's post-election commentary
Some Republican success, it is true, was down to recent shameless efforts to gerrymander the nation's congressional boundaries. But that does not fully account for their roughly 40-seat majority. Besides, 30 states now have Republican governors, though state borders cannot be gerrymandered.
The first assertion seems a bit dubious, as fewer votes were recorded for Republican House candidates than for Democratic ones, by a slim margin: 53.4 million to 53.9 million
. It may well be that it is state-by-state variation, rather than gerrymandering, which accounts for the discrepancy. But, certainly in a proportional-representation scheme we would face a Democratic majority (or perhaps a minority government in coalition.)
The second assertion just assumes that previous generations' border-drawing exercise is somehow natural. :) I can't find a quick source for the total population governed by Republicans vs. Democrats, but I wouldn't be surprised if the population count (or vote count) showed a closer or opposite result.
But talk of a "mandate" is generally nonsense to begin with, so I guess you might as well try to justify gridlock if that's your starting point.
|Saturday, November 10th, 2012|
|Saturday, November 3rd, 2012|
I happened to notice while looking at the Steam badges for Torchlight that more people have the badge for beating the final boss (Beast Slayer I, 17.8%) than for beating the penultimate boss (Only a Master of Evil, 16.0%.) This seems a fairly substantial discrepancy.
The only explanation I can think of is that some of the badges were introduced later? Or perhaps there's a way to merely avoid Alric instead of fighting him.
Or, I guess, the most likely explanation is just that Steam achievements are buggy. :)