Every day, choose a random MLB game to watch and tweet it as if it were meticulously selected. Data from Excel spreadsheets [[1](http://newballpark.org/wp-content/uploads/2013/09/MLBTravelGrid2014.xls), [2](http://dailybaseballdata.com/dbd/MLB_schedule_2014.xls)] or [an API](https://erikberg.com/api/methods/events), all of which are unofficial. ---- Take the attendance of a sports event, then subtract it from the event's venue, giving you X. Pick a random emotion E. Tweet something with the following pattern: "I feel [emotion E] for the [X] people [that didn't show up to | missed] [Team 1] beating [Team 2] [game score] at [Stadium S] [today | tonight]." ---- Take a Twitter timeline (follow some image-heavy accounts), and send all the images to Google Image Search. Scrape the suggested search for that image, and just post the text. Or maybe the synonyms for each word. Or maybe rhyming words for each non-article word. Or maybe antonyms: "The opposite of this image is...". ---- Every time I check into Foursquare and start off with "Reading" and then a book title + author name, post to my "Richard is Reading" Twitter account with a geo-located tweet. Maybe do an Amazon.com lookup of the title. More detail at [on my notes page](http://notes.justagwailo.com/mashups/foursquare). A levelled-up app would allow anybody to connect their 4sq and a Twitter account of their choice and post there. ---- Less a Twitter bot idea than a general (baseball-related) bot idea: 1. you check into a stadium in Foursquare and reference a specific game that you're attending 2. the bot looks up the starting lineup and picks a player to focus on for a whole inning. Weighted towards position players and against pitchers and especially the DH. 3. you watch that player and that player only. See how they prepare for each pitch, what they do while on the bases, etc. They never leave your sight. This won't work if you don't attend the game. TV broadcasts can't follow individual players. ---- https://twitter.com/search?q=from%3Ascanbc%20pedestrian&src=typd ScanBC is an amateur radio scanning community, and they tweet out any incident they hear involving emergency crews. Sometimes incidents involving collisions between pedestrians and vehicles cluster, so it would be interesting to build some stats. One could just search for 'from:scanbc pedestrian' and get raw counts, but it would be interesting to have a map of the incidents and see how they cluster. ---- Make a list of tweets to be posted anytime in the future. Set a timeframe of normal waking hours. Randomly select a tweet from the list. Randomly select the day and time to tweet them. Give me a reminder 5 minutes ahead of time and give me a y/n prompt. Alternatively, [a vigilance control](http://en.wikipedia.org/wiki/Dead_man's_switch#Vigilance_control). That is, stop tweeting if I haven't signaled that I'm alive/not comatose. Not like Buffer where time slots are pre-selected, and not like HootSuite where individual tweets are scheduled. It could use HootSuite to do the scheduling/tweeting, but the bot would have to choose the time randomly given the constraints above. ---- Get three random words from [Wordnik](http://www.wordnik.com/) and find the location associated with them at [what3words](http://what3words.com/), assigning the location to the tweet's metadata. Download a static map using [the Google Maps Static Map API](https://developers.google.com/maps/documentation/staticmaps/) and adorn the tweet with the map's image. ---- Using Foursquare's real-time API, when you check in on Foursquare, take the latitude and longitude of the venue and feed it to [what3words](http://what3words.com/). Capitalize the words in the response and say that you're at that location, but use the 4sq.com link so people can know where you really are if they click the link. ---- After a day or so, look at your previous day's tweets. Take the replies you made and look to see if there are any faves, retweets or replies to those tweets. If not, store in a database the name of the person who you replied to with the the tweet in question. At the end of the month, tally up the usernames you replied to and rank them. Send an email with a report of the replies that went ignored. The people with the most ignored tweets are ignoring you, so maybe you can spend the energy more usefully. Also, use the data to evaluate which replies got ignored. Maybe those aren't funny, insightful, etc. ---- A bot that follows some high-frequency accounts and RTs one tweet from the previous 24 hours from each account. The interval would be fuzzy, that is, 24 hours give or take. The Twitter account associated with it would be private and only followed by you. The bio would be up-front about what it's doing and whom it's doing it for, in case the person feels uncomfortable about it and decides to block it. ---- A bot that takes all your unfinished bot ideas and randomly selects one of them for you to work on, and tweets you. Maybe it looks at the people you follow (or the people that follow you) on Twitter, and says "Hey, @[random follower of you], the next time you see @[your twitter account], remind [gender] that to work on [project XYZ]!" to give some social oomph to it. Those followers would have to opt-in, but could only be people that you know. You'd have to have a machine-readable list of unfinished projects, maybe some of them have URLs like a GitHub repository. ---- Related to [a previous idea](https://thoughtstreams.io/sillygwailo/twitter-bot-ideas/#card-3855), take the tweets from last night and pick 50 of them at random. When you wake up, you'll have an email waiting for you with those 50 tweets. It would pick the tweets from where you left off if you have [a $25/month Tweet Marker](http://tweetmarker.net) account (I might be interested in a per-account price if it was in the low single digits per month, say $2/month) account, or just from a certain time if you don't. If I thought for a second you could reliably embed tweets using Twitter's JS, I'd suggest making them all embedded tweets so that Cards show up too. Set the Tweet Marker to the latest tweet at the time when the email was sent. You'll just have to be OK with missing the tweets not included in the email, safe in the knowledge that a computer looked deep within your soul and chose tweets based on an algorithm you half-understand. The bot/app would be called Tweets From Last Night. ---- Have you seen people post screenshots of text on Twitter, and highlight (by selecting the text) a section they want to call out? Here's an example: https://twitter.com/susie_c/status/450480207627972608 I thought maybe one could automate that process, and make it random. It would assume [PhantomJS](http://phantomjs.org/) is capable of selecting text as a person could with a mouse (I'm not sure it's possible to automate that part), and then take a 'screen cap' of just the text near and including the selection (I'm sure that's doable). You'd probably want to screencap a "Readability" version of the page, plus it would pick an article at random (somehow?). ---- I'm going ahead with https://thoughtstreams.io/sillygwailo/twitter-bot-ideas/3737/ I have the first half built, and the second half is to register the Twitter account. Then I need to figure out how to post a tweet with multiple images, one with the Google Map it creates, another with the first result from a Google Images search for the words it comes up with. Then associate the tweet with the location. Do I make the words hashtags? There's [a Wordnik Hackathon](http://tinysubversions.com/stuff/wordnik-hackathon/) coming up. Since I'll be traveling on the Friday, I won't participate directly, but I do hope to have something ready to show off by the time it rolls around. ---- A bot that randonly selects a brewery from Vancouver using GrowlerFill.ca's API (once it's made public) and randomly selects a beer (once the data is available) from that brewery then gets the Untapped rating for that beer and tweets a 'recommendation'.