What are some good metrics to measure increased productivity due to AI assisted coding?
Here are some that I can think of (not in order of importance:
1. Sprint velocity
2. Reduction of Bugs backlog
3. Number of PRs merged
4. Number of non trivial code reviews
Any additional that I missed?
Atlassian has some of the worst UX for user management, and it all seems intentional to trick customers into signing up for services they don't need.
For example, any new member added to JIRA is automatically added to other products (like Product Discovery etc) and those plans get upgraded to be paid plans. Mercifully, they downgrade the plans without charging, but this is super annoying. They also make it really hard to figure out how to remove a user from a product.
@Atlassian@Jira
Interesting. Could be a variety of reasons:
1. Wasn't there a study recently that indicated that it made only top developers a lot more productive, but other developers were actually less productive. If only top developers were more productive, their gains would be lost in the average.
2. Maybe it's enabling non-coders who aren't even using git.
@NTFabiano The people are conditioned by their society to feel emotion in a specific body part. So, these results are biased by that societal conditioning.
It would be interesting to also do some physiological measurements to validate this.
@NicoleLewisLCSW The map tells you what these emotions are, but how do you know which emotion you are feeling?
Best solution is a trained therapist, a substitute would be some online tool that can give you feedback from facial expression/audio/speech text emotions.
Indeed. Charging at home and not having to go to a gas station, and smell the fumes of gasoline on my hands is an unexpected perk I have been enjoying for 8+ years now. Not having to get pollution check done, nor 6 month oil change appointment, nor even brake pad replacement are bonus perks.
Aravind — some quick thoughts here from someone who’s lived inside the machine for a while.
Bloomberg isn’t valuable because it connects you to analysts. Sell-side coverage and access to management are “pay-to-play” through brokers — Bloomberg just aggregates it. Pricing data comes from the exchanges. Compliance isn’t native either; firms still run their own.
Bloomberg’s real edge is that it’s been around forever, has verified and structured decades of trading data, and offers a cleaner, more usable front end than anyone else. That’s how Mike won against Quotron and Cron, and why Thomson Reuters & Factset who tried competing against him on price couldn’t really make a dent.
You’re right though — there’s no fundamental reason a terminal should cost $30K/year. The moat is historical and UX-driven, not structural. There’s real room for disruption here for sure.
@amuse How is this a saving? Suppose someone didn't fill out a field inadvertently and got rejected. Wouldn't they try again filling in properly? You should report what percentage of claims are never tried again.