Skip to content

Five Features of Better Arguments – The Atlantic

Earlier this month, amid a particularly trying stretch on social media, I joked that I would soon be launching Ad Hominem, “a new journal of non-ideas founded in response to apparent massive popular demand. Issue One: You’re Trash.” While there never was a golden age of argument, the impersonal hyper-connectedness of the internet and the near omnipresence of our access to it have transformed the civic experience for those who remember something different.

“We have gotten habituated to penalty-free trashing of each other,” says Eric Liu, a former speech writer and policy adviser in the Bill Clinton administration.

Widespread frustration is understandable.

Still, Liu believes that “we don’t need fewer arguments today; we need less stupid ones,” a theory he first advanced in The Atlantic just prior to the 2016 presidential election, and that he continues to believe after launching and leading an effort to host more constructive civic exchanges all around the United States.

On Monday at the Aspen Ideas Festival, co-hosted by The Atlantic and the Aspen Institute, he sketched out what he now regards as best practices, drawn from insights gained through his work for the Better Arguments Project. The framework it has developed to guide local events is worth pondering.

Among its main tenets:

  1. Take Winning Off the Table: Rather than seeking victory, the goal should be truth-seeking, with a reinstitution of civility in service of achieving it. Participants are charged with arguing in order to better understand.
  2. Prioritize Relationships and Listen Passionately: As one audience member put it, the most constructive and rewarding arguments they’ve ever had involved people with whom maintaining a good relationship afterward was a high priority—an impetus for speaking and listening carefully.
  3. Pay Attention to Context: “One aspect of this concerns history,” Liu said. “Every fight we have today, about immigration, about taxes, about the minimum wage, is a recapitulation of one of those core American arguments—about liberty versus equality, about central government versus local control, or individual responsibility versus collective responsibility—and the history of civic debates in this country has something to teach us about how we can make our way through this conversations today. A second element is about emotion. If someone comes at you in an angry way, you have to adjust how you’re going to come back at them. And you have a choice about whether you’re going to mirror and double down or if you’re going to be the one to say, I’m gonna be the grown-up here and I’m going to deescalate—being emotionally intelligent about the patterns that we fall into.”
  4. Embrace Vulnerability: “Every one of us can relate to the feeling, ‘I didn’t start this, I’m not going to extend the olive branch.’ Extend the olive branch,” Liu said.
  5. Be Open: “You cannot possibly change another person’s mind,” Liu said, “if you’re not willing to have your own mind changed. You may be able to rack up debater’s points. But you won’t change their mind if they sense you aren’t willing to have your mind changed. It’s a matter of mindset but also ‘heart-set.’”

That isn’t the last word on how to argue. But anyone who embraces those five pieces of advice is almost guaranteed to have more constructive arguments.

via Five Features of Better Arguments – The Atlantic


Most of what you read about growing an Instagram audience is either irrelevant or not enough

Most of what you read about growing an Instagram audience is either irrelevant or not enough

“Post consistently. Use hashtags. Join an Engagement Pod. Rinse. Repeat.”

That’s basically the mantra, right? Every single blog post, YouTube video, or online course I read or watched on the Internet is either talking about all these things at once or going very vertical on one of them (the like of: “Definitive guide to Instagram Hashtags in 2018”).

I decided to use data analysis to check their validity.

I followed these rules, so to say, quite methodically for about six months using my company’s Instagram account, @uniwhere, and my own Instagram account, @giansegato. Uniwhere is an app for university students, so community building for us is key.
The following analysis is based only on Uniwhere’s account.

I varied the activities along the way – like stopping doing something while introducing something else; in this way, I could have the necessary degree of variation needed for a comparative analysis.

I then connected a Python notebook to the Instagram Graph APIs and to our own database, and tried to draw some conclusions.

TL;DR version: post consistency, hashtags exposure and engagement pods do not correlate at all with followers growth. You can benefit from them, but not in terms of audience development.

Let’s see the details.

The Data

As a starter, I’d like to point out which kind of data I was able to analyse. Here it is my final dataframe:

As you can see, I have a normalised set of metrics that are either coming directly from Instagram, or coming from our own internal ETL system.
Each row represents a day, starting from January 1st, 2018, and ending on June 10th, 2018.

I divided, for the sake of the analysis, the data into three buckets:


These are metrics that track explicit actions taken in order to grow the account:

  • did_post: whether in a particular day we posted a picture or not; it has a lagged version too, consistency: a trailing value that represents the degree of post consistency in the week before that day.
  • vgm: whether, that day, the post was followed through by a push with an engagement pod.
    Engagement pods push the engagement of a specific picture skyrocketing (hundreds of likes), and more often than not they allow you to have your picture featured in the top post section of a hashtag. They work pretty simply: there are groups of people that like each other pictures at a certain, agreed time. So, a lot of engagement and potential exposure.
    This feature has a lagged variable too, vgm_effect: again, a trailing value of 5 days to take into account the possibility that engagement pods have a shifted effect in the days after the push.
  • engagement: how much, on a particular day, we liked relevant pictures using hashtags of users we found on the platform. Another common, yet time-consuming, activity: you try to discover interesting pictures in a stream of pictures that are relevant to your nice, and you like them; the hope is that the receiver of the like visits your profile, likes it, and follows you.

Means to an end

These are the indirect results:

  • profile_views: how many users visited the account page in a given day
  • reach: how many unique users were exposed to our page in any manner
  • impressions: how many times one of our pictures was seen in a given day

Little spoiler: impressions and reach are perfectly correlated (1.0), so we could potentially use the two terms as synonyms, even though they mean two different things.


And finally, the true, final, KPIs that we are interested in:

  • followers_count: how many new users we got on a given day
  • website_clicks: how many times the website link we are featuring on the account page was clicked

In the end, it all looks like this:

Ok, not so useful, I get it.

Divide et imperat: let’s consider small comparisons before piecing together the whole picture.

Post Consistency

In my humble opinion, I believe that the big, everlasting rule of “post consistently and get new followers” is very misleading. Is it useful to post consistently? Well, it doesn’t hurt. For sure it’s critical for your existingaudience. But, looking at the hard evidence, it’s difficult to say that it helps growing it.
Let’s take a look at this plot:

Red, vertical lines identify the days in which we posted a picture (did_post = 1). Black, vertical lines are the days when not only we posted a picture, we also pushed it using an engagement pod (vgm = 1).

Looks like there’s no pattern at all between the two things. However, it’s kind of difficult to infer some interesting take from this image, other than the quite clear “engagement pods doesn’t help” (which is still extremely interesting).

Let’s plot the same data, but this time using post consistency (that – I remind you – is a moving-window sum, and therefore can be plotted as a continuous line):

As you can see between March and April (the red area), there looks like to be an interesting correlation between followers growth and post consistency: the more I’m posting (blue line), the more my followers are growing (green line).

However, later on, between April and May (black area), new followers growth was stagnating, while the consistency was record high. So… it doesn’t really look like that posting more, or more consistently, directly affects the follower’s growth, nor does using engagement pods.

Action no. 2: Be Trending with Top Posts

There have been moments when our page has been trending in massive hashtags streams. We’ve got some of our pictures that made it to be top post of a 20MN-stream or so. However, it didn’t bring us much engagement, and, probably even worse, basically no new followers at all because of it.

Take a look at the following plot. It’s showing the relationship between impressions and followers count.

As you can see, in general, they kinda move in the same direction (I guess). However, you can totally have audience growth with no impressions (the two red areas), as well as good impressions with no growth (black area).

So… nope. Having your picture seen around a lot doesn’t correlate much with followers growth.

Action no. 3: Engage Around Like Crazy

I’m just gonna show you this one.

New followers to engagement (small reminder: engagement is defined as a parameter between 0 and 1 that quantifies the intensity of the likes sessions to other users in the platform – how much we interacted in a given day with other users):

Website clicks to engagement:

Profile views to engagement:

It’s pretty much it.

When you like pictures around using the right hashtags, your profile views go up, and thus your followers count goes up as well, along with the website clicks.

That’s the only action that consistently proved to be a successful strategy to grow our Instagram audience.

Shall we cut the chase, please?

It’s time to have a look at the dataframe correlation matrix, which is basically a sum up of every possible plot I could potentially show you:

This picture extremely interesting. I’d need a few of these, from different accounts and different niches, in order to make a true generalisation, but as far as I can tell this is a pretty telling picture about how the Instagram game in 2018 works.
In other words, every possible strategy you might come up with should be covered in here, at least from a correlation standpoint.

My most interesting takes from this little purple guy:

  1. vgm_effect doesn’t correlate with anything. Like, really, anything. It barely correlates with profile_views, which is obvious because those people in your engagement pod actually have to visit your profile in order to like your last picture.
  2. did_post heavily correlates with impressions and reach, which, again, is obvious: when you post, you make impressions, when you don’t, you don’t. But no sign of correlation with follower_count.
  3. follower_count, on the other hand, is heavily correlated with engagementand profile_views, and somewhat correlated with website_clicks.
    As we already seen in the plots before, investing time in liking pictures around is the only way to actually make new followers: you like pictures, therefore your profile gets viewed, people are exposed to your gallery and then they decide if it’s worth following you.
    In the process of deciding whether to follow you or not, they might visit your website.
  4. To be fair, there’s some form of correlation between profile_views and impressions. However, I suspect that part of those profiles views that are coming after a spike of impressions (and therefore along with did_post) are your own followers visiting your page because you posted something. So, no new followers, but engaged existing audience.

Takeaways (ie. what I learnt from this)

In the end, I think that the main takeaways to be drawn here are the following:

1. Post only for your existing audience: do not expect your posts to generate more followers; not even if they “go viral

2. Use engagement pods only to boost your pics likes and have some social proof: do not expect engagement pods to be a magic wand to grow your audience

3. Like, like, like and like.

Final note

Why not take into account the follow-for-follow strategy (I’m following people around expecting part of them to follow me back)?
Well: I never did it, so I have no data about such strategy.

I believe it’s very successful in the short term, but I hate it, mainly because it conveys a very clear message of desperation, and you cannot scale it: either you keep every connection you make, and thus A. you hit the 7,500 limit fast, B. you truly look super desperate, or you unfollow every now and then some of your followers randomly, which is something that is quite dishonest, IMHO. So, nope, no data about it.

via Most of what you read about growing an Instagram audience is either irrelevant or not enough


Instagram Influencers Are Driving Luxury Hotels Crazy – The Atlantic


Three years ago, Lisa Linh quit her full-time job to travel the world and document it on Instagram, where she has nearly 100,000 followers; since then, she has stayed in breathtaking hotels everywhere from Mexico to Quebec to the Cook Islands. Often, she stays for free.

Linh is part of an ever-growing class of people who have leveraged their social media clout to travel the world, frequently in luxury. While Linh and other elite influencers are usually personally invited by hotel brands, an onslaught of lesser-known wannabes has left hotels scrambling to deal with a deluge of requests for all-expense-paid vacations in exchange for some social media posts.

Kate Jones, marketing and communications manager at the Dusit Thani, a five-star resort in the Maldives, said that her hotel receives at least six requests from self-described influencers per day, typically through Instagram direct message.

“Everyone with a Facebook these days is an influencer,” she said. “People say, I want to come to the Maldives for 10 days and will do two posts on Instagram to like 2,000 followers. It’s people with 600 Facebook friends saying, ‘Hi I’m an influencer, I want to stay in your hotel for 7 days,'” she said. Others send vauge one line emails, like “I want to collaborate with you”, with no further explanation. “These people are expecting five to seven nights on average, all inclusive. Maldives is not a cheap destination.” She said that only about 10 percent of the requests she receives are worth investigating.

Jack Bedwani, who runs The Projects, a brand consulting agency that works with several top hospitality brands, said that he’s close with the PR manager for a new hotel and day club in Bali. “They get five to 20 direct inquiries a day from self-titled influencers,” he said. “The net is so wide, and the term ‘influencer’ is so loose.”

“You can sort the amateurs from the pros very quickly,” Bedwani said.“The vast majority of cold-call approaches are really badly written. It sounds like when you’re texting a friend inviting yourself over for dinner—it’s that colloquial. They don’t give reasons why anyone should invest in having them as a guest.”

Some hotels report being so overwhelmed by influencer requests that they’ve simply opted out. In January, a luxury boutique hotel in Ireland made headlinesfor banning all YouTubers and Instagram stars after a 22-year-old requested a free five-night stay in exchange for exposure.

“If I let you stay here in return for a feature in a video, who is going to pay the staff who look after you? Who is going to pay the housekeepers who clean your room? … Who is going to pay for the light and heat you use during your stay? Maybe I should tell my staff they will be featured in your video in lieu of receiving payment for work carried out while you’re in residence?” the owner wrote on Facebook.

A post shared by Nicole Jayne White (@njwhite) on 

But to influencers themselves, this is a fundamental misunderstanding of the value exchange. Instagram has ballooned to more than 800 million monthly active usersmany of whom come to it for travel ideas, and influencers argue that the promotion they offer allow hotels to directly market to new audiences in an authentic way.

They’re not completely wrong. Most hotels acknowledge that there’s some benefit to working with influencers, it’s just that determining how to work with them—and manage their requests—is a challenge.

Some hotels, like the Ace and others, have attempted to standardize the process by requiring detailed Influencer application forms for discounts or free hotel stays. Others list influencer-specific contact addresses on their website. But the majority of hotels deal with influencer requests the old fashioned way, through an email to the hotel’s primary address. Many influencers use an email template that they customize for each property when requesting a stay.

Hotels evaluate influencers on several criteria, trying to sift through an enormous amount of BS. “We have quite a strict process,” said Jones. “We look at engagement more than anything else … We have to filter out influencers who have basically bought bots. There’s a lot of those these days.”

Laurie Hobbs, director of public relations and marketing at Ocean House Management, a resort management company that operates several boutique hotels in Rhode Island, said the hotel keeps a database of trusted influencers who it has partnered with before and can rely on when launching new products—recently, a customized Lilly Pulitzer suite. When new influencers approach one of their properties, Hobbs and her team take a close look at the influencer’s follower count and aesthetic to make sure it’s the right fit.

Joe Miragliotta, a men’s lifestyle and travel blogger who says he’s stayed in hundreds of hotels as an influencer, hates the fact that so many ham-handed wannabes are giving influencers a bad name. “You’ve seen the bad emails like, ‘Can you believe this influencer reaching out to me for a free stay!?’ That’s a bad thing. That makes us all look bad,” he said.

Miragliotta said having a clear pitch and meaningful deliverables can make all the difference to a hotel brand manager. “Having a one-sheet is really nice,” he said. “Have your demographics on lockdown. Have an elevator pitch. Know your audience … If you don’t know your audience, brands don’t know you. You could have 100 million followers, but they won’t know who you’re marketing to.”

Linh said that while hotels are still trying to figure out the return on working with influencers, it can be helpful to provide more than just social media posting. “We can film something for their website, or provide imagery,” she said. “They can save money by hiring an influencer vs hiring professional photographers or videographers.”

Other influencers have gotten even more creative with the services they offer. Zach Benson, who owns a network of travel Instagram accounts and who says he has gotten more than 200 nights for free over the past year and a half, touts his background in digital marketing when he approaches hotels. Along with the traditional Instagram posts and Stories, Benson offers to work with a hotel’s digital marketing arm to improve the brand’s in-house social media accounts.

“We really want to help people and make their companies and hotels better,” he said. “We know that just doing a couple Instagram posts for them isn’t really going to help them that much.” During his travels, Benson hosts boot camps for hotel social media teams, where he trains employees on things like Facebook ads and Instagram promotion.

“I just think a lot of the influencers have entitlement mentality,” Benson said. “A lot of them think about giving the bare minimum.”

Bedwani said that it’s critical that hotels set explicit terms in their deals with influencers. “I know a major brand that opened up and flew in a plane full of influencers,” he said. “Three quarters of them didn’t even post. It was a major fail from their team.”

Jones, meanwhile, said the Dusit Thani Maldives has all but ceased working with fashion influencers after she discovered that many simply wanted a pretty backdrop for their swimsuit shots.

“10 different bikini pictures a day on the beach is great for the bikini company,” she said. “But you can’t even tell where it’s taken. It could be anywhere in the Maldives.”

Some of these issues can just be a miscommunication. Miragliotta said he’s invested in making clients happy—but hotels need to make sure they’re organized and prepared for influencer stays.

“I went to one Mexico resort and three different people were giving me different hashtags and handles,” he said. “I was with five other influencers and we were excited to post, but there was limited wifi. If you don’t have the simplest things ready for us then that makes it difficult to produce the content you need, or do it correctly.”

Natalie Zfat, a social media consultant and influencer who has partnered with hotel chains like Marriott and InterContinental, said that at the end of the day most bloggers and influencers are just businesspeople.

“Could you think of any other business universe where it would be frowned upon for someone to reach out to a potential client and offer them an opportunity? You’d never see Coke call out an ad sales person for calling them up and sharing their rates.”

via Instagram Influencers Are Driving Luxury Hotels Crazy – The Atlantic

The surest way of concealomg from others the boundaries of one’s own knowledge is not to overstep them.

Giacomo Leopardi

Mr. Rogers’s Simple Set of Rules for Talking to Kids – The Atlantic

For the millions of adults who grew up watching him on public television, Fred Rogers represents the most important human values: respect, compassion, kindness, integrity, humility. On Mister Rogers’ Neighborhood, the show that he created 50 years ago and starred in, he was the epitome of simple, natural ease.

But as I write in my forthcoming book, The Good Neighbor: The Life and Work of Fred Rogers, Rogers’s placidity belied the intense care he took in shaping each episode of his program. He insisted that every word, whether spoken by a person or a puppet, be scrutinized closely, because he knew that children—the preschool-age boys and girls who made up the core of his audience—tend to hear things literally.

As Arthur Greenwald, a former producer of the show, put it to me, “There were no accidents on Mister Rogers’ Neighborhood.” He took great pains not to mislead or confuse children, and his team of writers joked that his on-air manner of speaking amounted to a distinct language they called “Freddish.”

Fundamentally, Freddish anticipated the ways its listeners might misinterpret what was being said. For instance, Greenwald mentioned a scene in a hospital in which a nurse inflating a blood-pressure cuff originally said “I’m going to blow this up.” Greenwald recalls: “Fred made us redub the line, saying, ‘I’m going to puff this up with some air,’ because ‘blow it up’ might sound like there’s an explosion, and he didn’t want the kids to cover their ears and miss what would happen next.”

The show’s final cuts reflected many similarly exacting interventions. Once, Rogers provided new lyrics for the “Tomorrow” song that ended each show to ensure that children watching on Friday wouldn’t expect a show on Saturday, when the show didn’t air. And Rogers’s secretary, Elaine Lynch, remembered how, when one script referred to putting a pet “to sleep,” he excised it for fear that children would be worried about the idea of falling asleep themselves.

Rogers was extraordinarily good at imagining where children’s minds might go. For instance, in a scene in which he had an eye doctor using an ophthalmoscope to peer into his eyes, he made a point of having the doctor clarify that he wasn’t able to see Rogers’s thoughts. Rogers also wrote a song called “You Can Never Go Down the Drain” because he knew that drains were something that, to kids, seemed to exist solely to suck things down.

In 1977, about a decade into the show’s run, Arthur Greenwald and another writer named Barry Head cracked open a bottle of scotch while on a break, and coined the term Freddish. They later created an illustrated manual called “Let’s Talk About Freddish,” a loving parody of the demanding process of getting all the words just right for Rogers. “What Fred understood and was very direct and articulate about was that the inner life of children was deadly serious to them,” said Greenwald.

Per the pamphlet, there were nine steps for translating into Freddish:

“State the idea you wish to express as clearly as possible, and in terms preschoolers can understand.” Example: It is dangerous to play in the street. ​​​​​​
“Rephrase in a positive manner,” as in It is good to play where it is safe.
“Rephrase the idea, bearing in mind that preschoolers cannot yet make subtle distinctions and need to be redirected to authorities they trust.” As in, “Ask your parents where it is safe to play.”
“Rephrase your idea to eliminate all elements that could be considered prescriptive, directive, or instructive.” In the example, that’d mean getting rid of “ask”: Your parents will tell you where it is safe to play.
“Rephrase any element that suggests certainty.” That’d be “will”: Your parents can tell you where it is safe to play.
“Rephrase your idea to eliminate any element that may not apply to all children.” Not all children know their parents, so: Your favorite grown-ups can tell you where it is safe to play.
“Add a simple motivational idea that gives preschoolers a reason to follow your advice.” Perhaps: Your favorite grown-ups can tell you where it is safe to play. It is good to listen to them.
“Rephrase your new statement, repeating the first step.” “Good” represents a value judgment, so: Your favorite grown-ups can tell you where it is safe to play. It is important to try to listen to them.
“Rephrase your idea a final time, relating it to some phase of development a preschooler can understand.” Maybe: Your favorite grown-ups can tell you where it is safe to play. It is important to try to listen to them, and listening is an important part of growing.
Rogers brought this level of care and attention not just to granular details and phrasings, but the bigger messages his show would send. Hedda Sharapan, one of the staff members at Fred Rogers’s production company, Family Communications, Inc., recalls Rogers once halted taping of a show when a cast member told the puppet Henrietta Pussycat not to cry; he interrupted shooting to make it clear that his show would never suggest to children that they not cry.

In working on the show, Rogers interacted extensively with academic researchers. Daniel R. Anderson, a psychologist formerly at the University of Massachusetts who worked as an advisor for the show, remembered a speaking trip to Germany at which some members of an academic audience raised questions about Rogers’s direct approach on television. They were concerned that it could lead to false expectations from children of personal support from a televised figure. Anderson was impressed with the depth of Rogers’s reaction, and with the fact that he went back to production carefully screening scripts for any hint of language that could confuse children in that way.

In fact, Freddish and Rogers’s philosophy of child development is actually derived from some of the leading 20th-century scholars of the subject. In the 1950s, Rogers, already well known for a previous children’s TV program, was pursuing a graduate degree at The Pittsburgh Theological Seminary when a teacher there recommended he also study under the child-development expert Margaret McFarland at the University of Pittsburgh. There he was exposed to the theories of legendary faculty, including McFarland, Benjamin Spock, Eric Erikson, and T. Berry Brazelton. Rogers learned the highest standards in this emerging academic field, and he applied them to his program for almost half a century.

This is one of the reasons Rogers was so particular about the writing on his show. “I spent hours talking with Fred and taking notes,” says Greenwald, “then hours talking with Margaret McFarland before I went off and wrote the scripts. Then Fred made them better.” As simple as Mister Rogers’ Neighborhood looked and sounded, every detail in it was the product of a tremendously careful, academically-informed process.

via Mr. Rogers’s Simple Set of Rules for Talking to Kids – The Atlantic

tmux on AWS

tmux on AWS
Part 1. Background
Part 2. Getting Set Up
Part 3. Using tmux
Part 4. Use tmux, Exit AWS, Log back in to AWS
Part 5. tmux: Summary of Primary Commands

Part 1. Background
Lets you tile window panes in a command-line environment.
This in turn allows you to run, or keep an eye on, multiple programs within one terminal.
🔑 With tmux, you can leave scripts running for a while, and it doesn’t matter if the terminal closes or you lose your internet connection for a moment; the script is running in the background
Tmux: A Gentle Introduction to tmux by Alek Shnayder

Getting In & Getting Out
Managing Panes
Custom Themes
Part 2. Getting Set Up
Step 1: A Background
Note 1: I am using fastai deep learning AMI
Note 2: tmux is already installed.
Note 3: On the fastai AWS AMI, tmux mouse mode is enabled, so hold down shift while selecting to copy to your local clipboard.
Note 4: If tmux is not installed:

Mac: brew install tmux
Linux: sudo apt-get install tmux
see the Resource above for instructions
Step 2: Log into AWS Instance
Note 1: Log into AWS Console and check that instance is running
Note 2: Ensure I am in the appropriate directory

my syntax for logging in

ssh -i aws_fastai_gpu.pem ubuntu@ -L8888:localhost:8888

my example

ssh -i aws_fastai_gpu.pem ubuntu@ -L8888:localhost:8888
The authenticity of host ‘ (’ can’t be established.
ECDSA key fingerprint is SHA256:RoHkdmFaS+2/w/9CGncGb4cPO3lUutStxQ7BACCzopI.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added ‘’ (ECDSA) to the list of known hosts.
Welcome to Ubuntu 16.04.3 LTS (GNU/Linux 4.4.0-1039-aws x86_64)

* Documentation:
* Management:
* Support:

Get cloud support with Ubuntu Advantage Cloud Guest:

5 packages can be updated.
0 updates are security updates.

(fastai) ubuntu@ip-172-31-10-243:~$
Update packages
sudo apt-get update

my example

(fastai) ubuntu@ip-172-31-10-243:~$ sudo apt-get update
Hit:1 xenial InRelease
Hit:2 xenial-updates InRelease
Hit:3 xenial-backports InRelease
Hit:4 xenial-security InRelease
Hit:5 xenial InRelease
Reading package lists… Done
(fastai) ubuntu@ip-172-31-10-243:~$
Part 3. Using tmux
See what version of tmux is running
tmux -V

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux -V
tmux 2.1
See what tmux sessions are running
tmux ls

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
error connecting to /tmp/tmux-1000/default (No such file or directory)
Note: right now, I have no tmux sessions running.

tmux: start a tmux session
Note: A green bar will appear at the bottom

Confirm tmux sessions are running
tmux ls

tmux commands
hit ctrl and b at the same time and then let go
then type subsequent command
example for creating two stacked windows
ctrl and b (at same time)

tmux help commands
ctrl+b ?

tmux: create a second window (horizontal)
ctrl+b ”

tmux: create a third window (vertical)
ctrl+b %

Note: type ctrl+b then let go, and then type %

tmux: navigate between windows
ctrl+b [then arrow up/down/side]

tmux: MAGIC command
🔑 This is the key prompt for using tmux.

tmux: kill current pane
ctrl+b x

tmux: detach session
This will detach the current session and return you to your normal shell. You can exit AWS with tmux running in background, and ssh into the instance again.
ctrl+b d

tmux: list sessions from normal shell
tmux ls

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
0: 1 windows (created Sat Nov 11 15:31:41 2017) [159×38]
(fastai) ubuntu@ip-172-31-10-243:~$
tmux: connect back to tmux session from normal shell
tmux a -t 0

tmux: kill tmux server, along will ALL sessions
tmux ls
tmux kill-server
tmux ls

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
0: 1 windows (created Sat Nov 11 15:31:41 2017) [159×38]
(fastai) ubuntu@ip-172-31-10-243:~$ tmux kill-server
(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
no server running on /tmp/tmux-1000/default
(fastai) ubuntu@ip-172-31-10-243:~$
Part 4. Use tmux, Exit AWS, Log back in to AWS
Start a tmux session on AWS
tmux ls
tmux ls

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
no server running on /tmp/tmux-1000/default
(fastai) ubuntu@ip-172-31-10-243:~$ tmux
[detached (from session 0)]
(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
0: 1 windows (created Sat Nov 11 15:52:31 2017) [159×38]
(fastai) ubuntu@ip-172-31-10-243:~$
(fastai) ubuntu@ip-172-31-10-243:~$ exit
Connection to closed.
Log back into AWS & Confirm tmux session is still running
ssh -i aws_fastai_gpu.pem ubuntu@ -L8888:localhost:8888
tmux ls

my example

(fastai) ubuntu@ip-172-31-10-243:~$ tmux ls
0: 1 windows (created Sat Nov 11 15:52:31 2017) [159×38]
(fastai) ubuntu@ip-172-31-10-243:~$
Note: The tmux session is still running!

Part 5. tmux: Summary of Primary Commands

via fastai_deeplearn_part1/ at master · reshamas/fastai_deeplearn_part1 · GitHub

Moral maxims are surprisingly useful on occasions when we can invent little else to justify our actions.

Alexander Pushkin