HNHacker News
TopNewBestAskShowJobs

trjordan

5,143 karma · joined October 8, 2008

Catch the decisions your agent didn't tell you about https://tern.sh

tr at tern dot sh

submissionscomments
trjordan··on Facebook Knows How to Track You Using the Dust on Your Camera Lens
They're basically the Google Photos team, now. Have you seen the ads about face recognition and auto-sharing? They've been working on the problem of photo sharing ever since they figured out that people didn't want to share contact info with Bump, they wanted to share photos.
trjordan··on Talent Is Everywhere, Opportunity Is Not
This is true, but one of the important inputs to socioeconomic asymmetry is race and gender.

It's hard for your parents to escape poverty if the police profile and arrest them. Or if your mom could have been promoted faster than your dad, but wasn't because of a bad company culture.

trjordan··on The future of Kubernetes is virtual machines
Config blast radius, hardware sizing, networking partitions, disaster recovery, dev/prod separation -- lots of good reasons to have multiple clusters!

As a gut check, most of these reasons apply to AWS AZs as well. If your Kubernetes strategy calls for more than one per region per AWS account, it means that you, organizationally, don't trust containers as much as you trust VMs on AWS. And right now, you're probably right to do so.

trjordan··on The future of Kubernetes is virtual machines
This is, like all good HN articles, technically correct and practically incorrect.

It is correct that containers leak, and people know this. Multi-cluster strategies are real, and they shouldn't be. It should be OK to have one big cluster[1]. Until Kubernetes fixes this, there will be some friction to adopt it, based on real use cases like untrusted code and noisy neighbors.

It is incorrect because users (e.g. non-infrastructure engineers) don't know or care about the precise definition of containers and VMs are. The point of "containers" is that I can define something that acts like an operating system from the ground up, and it builds quickly and runs quickly in production.

Kubernetes doesn't win by forcing users to think about VMs. Kubernetes wins by adopting a VM standard that can be built by Dockerfiles. Infra engineers will love it.

But besides them? Nobody will care, because Docker for Mac will look the same.

[1] Maybe 1 cluster per region? There's a whole fascinating topic that starts with the question "when building a PaaS, do you expose region placement to devs?" The answer implies a ton of stuff about what exactly it's reasonable to expect from a PaaS and how much infrastructure your average dev has to know.

trjordan··on Bias in student evaluations of teaching
Did you read the linked papers? They go into a lot of detail about how they did it.
trjordan··on YouTube CEO calls EU’s proposed copyright regulation financially impossible
YouTube Vimeo is more like 1% / 99%. By almost anybody's definition, that's winner take all.

Vimeo aims for $100mil this year: https://www.reuters.com/article/us-iac-interactive-vimeo/iac...

YouTube Could be a $15 billion business this year: https://www.fool.com/investing/2018/02/18/youtube-could-be-a...

trjordan··on Linus Torvalds apologizes for his behavior, takes time off
"Given enough eyeballs, all bugs are shallow." - Linus Torvalds

A rude leader drives people away from the organization. It is directly counterproductive.

trjordan··on Chrome 69: “www.” subdomain missing from URL
It's technically correct that www.example.com and example.com are different domains and can serve different stuff.

So are https://example.com and http://example.com.

If you serve materially different content based on small differences, that's user-hostile, and common tools have no obligation to support you. Chrome shouldn't cater to sites that change behavior based on a www prefix, because the vast majority of users expect those sites to be the same. Hiding the prefix won't change their mind.

trjordan··on Cargo ships that sink when their cargo suddenly liquefies
It would cost a bunch of money but go part of the way towards fixing the problem.

But if you're going to spend the money, maybe you could entirely fix the problem. You'd have to run the numbers!

trjordan··on Holloway Guide to Equity Compensation
At anything other than really-low-valuation-seed-stage, the cost to purchase is going to be material. Since 409(a) is basically just 1/5th of the last valuation, at $5mil valuation, 1.0% equity will cost you $10k to exercise. If you can burn $10k without thinking about it on startup salary, this advice isn't for you.

If you wait, you get substantially more information. A company that makes it to a Series A has a much higher chance of succeeding, and you'll typically know that in your typical tenure of 1-3 years as a junior dev.

If you want to spend $10k cash on buying startup equity, at least call yourself an angel investor and go find a company that's willing to issue you a note that converts to preferred stock. It's too easy for a company to have a couple of missteps and wipe out most of the common stock value, even much later.

trjordan··on Holloway Guide to Equity Compensation
You probably shouldn't early exercise, which makes the question of 83b moot.

You are investing a couple years of your life in this. That's probably investing a couple hundred thousand dollars of opportunity cost vs a job at large tech company.

You have plenty invested. You'll do well if they do well. Spend your liquid investments in something diversified.

- sincerely, a moderately lucky person who has joined 3 companies before employee 15

trjordan··on Systemisers are better at maths
"A 2010 analysis of autism diagnoses in California did not find that autism clustered preferentially around areas rich in IT industry. Instead, it found that clusters tended to occur in areas where parents were older and educated to a higher level than were parents in surrounding areas."
trjordan··on Do You Know What Lightning Really Looks Like?
Also: lightning is bright and fast. It's difficult to see the branches that aren't as bright as the brightest few.
trjordan··on Former Equifax Manager Charged with Insider Trading
I disagree with the notion that the penalty for white-collar crime should be strictly proportional to how good you are at that brand of crime.

In a world where people occasionally steal $100m, that implies a maximum sentence of ~1 year for stealing $1m.

trjordan··on Ask HN: Pros and cons of working at a startup in 2018?
Comp is a big one, discussed elsewhere.

The other professional problem that well-run startups have is that it's easy for early employees to feel like they're held to a different set of standards than the founders for professional growth. Essentially, founders are given chances to succeed in roles they never could otherwise, and early employees miss all the coaching and context that might help them succeed in similar roles.

Let's imagine I was a 26-year-old developer interested in startups. I have two paths:

- I found a company. I have the chops to hack something together with my cofounder, and maybe we make something of it raise a seed. We hire another couple engineers, and with a couple big logos in a good market, we raise an A. At this point, I've learned a ton about what makes a business work from other founders and I see (far too clearly) the gaps in my own knowledge. I hire some marketers / salespeople to work for me, and we keep making things work. At some point, maybe I choose to hire a professional CEO, and I stay on as CTO. I still have a lot of leverage in the company. I still go to board meetings.

- I join a company as the first employee, after they've raised a seed. I have good conversations with the other 26-year old founders over beer after work. I jump on the occasional phone call with customers and crank out features that they like based on that insight. We raise our A, though I'm still a software engineer since we only have 10 employees that all report to one of the founders. I quit seeing customers, because we hired a sales rep. We decide to hire management, and I'm considered for a line manager job, but we ended up hiring somebody externally. (She was awesome, glad we hired her, but I'm bummed I'm not reporting to the CEO anymore.) By the time we're 40 people, I don't talk to the founders much anymore, and I don't really understand what our bigger customers want. I don't have the influence I once did, and the codebase is too big for me to brute force that leverage through productivity. I start to look for another job.

There's nothing wrong with either story, but in the founder case, I was given a ton of chances and support. As an employee, I was given, at best, the support of a single manager while the company grew around me. This is fair, in some sense, but compared to the opportunities offered to the founders, it's far behind.

trjordan··on GitHub and Open-Source Is a Boon for the Underprivileged
Thinking about Github-as-resume as strictly positive or uselessly noisy probably means you have an overly simplistic model for evaluating candidates.

I'll share a model that's worked for me in the past. Before you even write the job description, define the 3-7 traits you need to hire against. These can be as fuzzy as "gets shit done" or as specific as "5+ years marketing SaaS to $100k+ accounts" or "is effective at resolving incidents while on-call." You need to be able to rate people on this. Binary is fine ("yes they can get shit done"). 3-4 levels is good (bad / good / total superstar).

Don't go overboard with granularity: you need to fit this all in your head for a couple candidates at a time when you compare them. As you go through your process, try to get good information about at least one of these areas at each step.

My favorite thing to do is, between in-person interviews, be able to tee up the next interviewer with "I got mixed signals about communication style. Can you dig in?" When you wrap up, make sure you cover all the boxes. Don't get smitten with a candidate with amazing experience who can code circles around everybody else who's never worked in a 25+ person team (assuming that's what you agreed was important). When you're ready to make a decision, you now have a reasonable set of trade-offs to make. I tend to create rubrics where solid scores in 3 of 5 categories is a Hire, to allow us to compare different types of candidates.

Ok, back to GitHub.

GitHub can give you VERY STRONG SIGNALS for some of these issues. It's great to be able to know up-front that somebody is comfortable working with teams they don't personally know well (typical in big open-source projects). The convese to that is that I generally expect some strong signal up front from any candidate. If not GitHub, give me a respected company on your resume. A recommendation from a mutual friend. A well-written cover letter. When folks complain about 100s of resumes, it's because the problem is this lack of signal.

An aside: it's on you to allow folks with non-traditional backgrounds to succeed at this step. Write a better job req, or go email likely candidates. Diversity requires multiple ways in. Once you're a bit further, this cuts the other way. If "good communication in pull requests" is important to you, don't spend any time here with candidates who have strong Githubs.

You already have all the data you need. Spend your time elsewhere. There is no extra credit. GitHub can help you a bunch when evaluating candidates, but it's crucial to stay away from this notion that _anything_ during an evaluation that looks like a silver bullet. You always have to do the work of interviewing well, and it's always specific to your company.

Finally: if you have a process that allows you to fully evaluate a candidate by reading their public GitHub, you're almost certainly hiring for the wrong things.

trjordan··on Alphabet shareholders reject diversity proposal backed by employees
Try Googling "diversity performance teams."

"Why Diverse Teams Are Smarter" - HBR [1]

"Diversity and Work Group Performance" - Stanford [2]

"New Research: Diversity + Inclusion = Better Decision Making At Work" - Forbes [3]

There's a drumbeat of this stuff, both old and new, that comes out constantly.

[1] https://hbr.org/2016/11/why-diverse-teams-are-smarter

[2] https://www.gsb.stanford.edu/insights/diversity-work-group-p...

[3] https://www.forbes.com/sites/eriklarson/2017/09/21/new-resea...

trjordan··on Alphabet shareholders reject diversity proposal backed by employees
White Americans includes hispanic and non-hispanic.

In particular, though, the only over-represented group is asians: 35% vs 5%.

trjordan··on Ask HN: Best practices for onboarding new employees?
Find a way to listen to what's happening around you without being overly distracting. Bias towards current projects. Here's a couple things I've done in the past:

- Sit on customer calls

- Read whatever approximates the "done recently" engineering tasks: git commits, Jira, sprint boards, whatever.

- Show up to meetings you weren't invited to.

- Take 6 different people out for coffee and ask them all the same 6 questions. Whatever you want to know.

- Sit on customer calls

- Set up the product and try to do something with it.

- Better yet, set up the product and try to do a thing a customer is trying to do.

- Did I mention sit on customer calls?

Once you know what people are working on today, it's easier to pick up and help them. Then you can start in on your own initiatives with credibility and support from the people you've helped.

trjordan··on USA needs law 'a lot like GDPR' says Salesforce CEO Marc Benioff
Salesforce, roughly speaking.
trjordan··on How to Market Your SaaS Product
Assumptions that apply to this author, but may not apply to your business:

- The pricing is $15 / user a month. This is a small purchase for businesses.

- The product is in a crowded market. Comparison shopping is the norm.

- The target is SMBs. Simplicity / ease-of-use is big for this market, vs larger businesses.

- The product is in a mature market. People have well-defined expectations of what this software should do.

- The audience is customer support, or perhaps generalists / founders.

- Testing the software in "production" is hard, because it requires you to answer support tickets through it. It may be impossible to get full value from this product without committing to a trial, which implies a certain amount of investment.

The conclusions in the post are highly dependent on the specifics of this product and market, so take them with a grain of salt. If you're making a better CI/CD tool, or a better analytics dashboard, you may find that an entirely different approach works for you.

trjordan··on Interns with toasters: how I taught people about load balancers
Outlier detection also fixes this, without having to implement health checks that replicate the whole functionality (complexity) of the rest of the service.

10 500s in a row? Go to the timeout chair until you get better.

trjordan··on How founders can write a quality blog post in 1 hour
Jeez, no, but there's no way I'm fixing it now!
trjordan··on How founders can write a quality blog post in 1 hour
It depends on your audience.

If you're trying to build an audience based on pithy advice in a well-known space, there are a lot of posts like this that apply. Focus your ideas, write about one thing, make it relatable, bold a couple random words, and Bam! Traffic!

If you're writing about something technical, complex, or sophisticated, it's worth spending more time to get it write. If you're writing for developers, include code and make sure it runs. If you're writing about hard topics, find research to back up your conclusions and build a real narrative (Priceonomics does this well).

Especially as a founder, your goal should be to write the definitive piece on whatever your topic is. As somebody starting a company on that topic, you are legitimately one of the foremost experts on this topic, so people want to hear your thoughts! Keep all those other rules in mind, buy don't sell yourself short by writing half-hearted pieces that sort of rehash the same topics.

For example, where I work, we think that traffic management (routing, proxies, service mesh, all that rot) is a better way to do things if you've gone to microservices. One example is releasing software. So we wrote one article on it (in two parts). It's long, it's technical, and it has been so, so successful. We don't try to reinvent the wheel on this topic. We just point people to this post.

https://blog.turbinelabs.io/deploy-not-equal-release-part-on...

trjordan··on Text Embedding Models Contain Bias
One problem is the world has layered effects of bias.

Imagine a simple system where people are only evaluated for credit cards and mortgages, and businesses can only offer one of these services. Some people are systematically denied credit cards. Now they have to have more cash for daily needs, which makes them objectively worse as home owners, so the mortgage loaners pick up on this fact and start systematically denying them mortgages.

There's no business opportunity here. For either lender, the prospects in this group are worse, only because the other lenders think so. Breaking the barricade is much harder than just starting a new lender or either type.

trjordan··on Zuora S-1
Consider a company that used to sell stuff by the unit. Boxed software, DVDs, whatever.

There is a new world of their product, and even though it's 2018, they're not in it. It might feel like Spotify has won, Netflix has won, and so on, but in other verticals, subscriptions still aren't the norm.

Remember when Dollar Shave Club started selling razors by the month and got bought for a billion dollars? P&G needed to find a way to transition to subscriptions, and they paid up for it.

The subscription economy is a real trend, and it's not something you can just convert to overnight if you have a business that's a going concern. You have to figure out how to put out some new offerings, and the way people pay for these new subscriptions have to work with your existing business processes.

Zuora sells to those companies. Companies that want to sell subscriptions, but can't throw the baby out with the bathwater. They want to work with a company that understands that subscriptions are the new important thing, but not the only thing.

trjordan··on Zuora S-1
My experience with Zuora is that it's a feature-complete payments subscription that's an unholy pain to work with. We chose them over Stripe before Stripe has ACH.

My favorite part is that 25% of their revenue is professional services, and their professional services are sold essentially at-cost. You're going to need somebody to help you get set up, but at least the pricing on that extra headcount is fair :)

They're after large customers, and this is what they get.

"As a substantial portion of our sales efforts are increasingly targeted at large enterprise customers, our sales cycle may become increasingly lengthy and more expensive, we may encounter still greater pricing pressure and deployment and customization challenges, and we may have to delay revenue recognition for more complicated transactions, all of which could adversely impact our business and operating results."

trjordan··on The Business of SaaS
The benefit enormously from marketing, but they don't benefit from high-visibility brand marketing.

Trade shows, performance marketing, affiliates, content marketing, retargeting, account-based ads ... those all work well.

trjordan··on Ask HN: Is there server-side software that we are missing in 2018?
10 years ago, I learned about the idea of elastic infrastructure. Services with load balancers in front of them, hosts that come and go easily, a structured way to communicate between services.

At the time, that was nginx + manual autoscaling, then it was ELBs and autoscaling groups, now it's kubernetes and containers, maybe hosted. It's still not there.

I'm excited about the software that makes that operable at scale. It seems like a service mesh is a good idea. It seems like mutual security between services is a good idea. It seems like storing routing configuration in a separate control plane that is executed in a data plane like Envoy is a good idea.

There's a bag of software at CNCF that's loosely organized around this, but I don't think the "just deploy some code, it can scale and you can have tons of services doing that with good visibility and operability" is quite there. I'm really exicted about Envoy, but I don't think there's a good control plane for it. I'm part of a company that's working on a commercial implementation (turbinelabs.io), and Istio is in a similar space.

There's still work to be done!

trjordan··on Automerge: JSON-like data structure for building collaborative apps
This is cool! I've been thinking about this a lot in a similar domain (JSON / YAML for configuration), and I've come to the conclusion that anything approaching manual merge is pretty much the worst thing ever.

The domain I'm specifically thinking of is managing routing rules, like nginx configs or control-plane configs for Envoy. It's really tempting to simplify it down to a config file, then you version-control it, then you get merge conflicts, and all of a sudden a bad merge makes 10% of your services unroutable. Which is hilariously bad.

Having an auto-merge for those sort of configs seems like the only reasonable way to do it. Escalating conflicts up, from easy merge, to trying an automatic strategy, to surfacing conflicts to the user in a domain-specific way, seems like the only way to do it!

← PreviousPage 6 of 20Next →