My skip manager started asking for weekly status reports so my manager started pulling Jira reports and feeding them to ChatGPT. It turns out that my skip manager was using Copilot to summarize those reports into basically what you could get directly from Jira.
"Recent outage was due to a retry loop for the Foo API exceeding rate limits. We're implementing a backoff algo"
Sender, via ChatGPT:
Hi,
I wanted to provide more context regarding the recent outage.
The issue was triggered by a retry loop in the Foo API integration. When the API began returning errors, our system initiated repeated retry attempts without sufficient delay, which quickly exceeded the rate limits imposed by the API provider. As a result, requests were throttled, leading to degraded service availability.
To address this, we are implementing an exponential backoff algorithm with jitter. This approach will ensure that retries are spaced out appropriately and reduce the likelihood of breaching rate limits in the future. We are also reviewing our monitoring and alerting thresholds to detect similar patterns earlier.
We’ll continue to monitor the system closely and share further updates as improvements are rolled out.
Best regards,
Receiver, via ChatGPT:
"The outage was caused by excessive retry attempts to the Foo API, which triggered rate limiting and degraded service. To prevent recurrence, exponential backoff with jitter is being implemented"
> Ultimately a lot of this generative tech stuff is just counterfeiting extra signals people were using to try to guess at interest, attentiveness, intelligence, etc.
> So yeah, as those indicators become debased, maybe we'll go back to sending something [...] all boiled down to bullet points.
It's already started, with AI resume spamming on one side and the associated AI assisted screening on the other. "Deep research" generated funding requests on one side and associated AI generated funding request summarization on the other.
This is the kind of thing my manager and skip manager are up to. Most of our jira tickets now are written as a random jumble of ideas, fed into internal proprietary LLM and then turned into jira description and acceptance criteria. Totally pointless
This guy is quite possibly going to end up looking stupid when something goes wrong and it turns out he lied about having thought about it. I hope he is as clever as he thinks he is at anticipating what will go wrong in the future. Fires and whatnot do happen. Even AWS us-east-1 has experienced outages.
The author jokes, but almost everybody's DR plan (at least, the DR plans motivated by regs like SOC2 --- which I believe are most DR plans) are worse than what you'd get from an LLM. An LLM can at least take some input and craft something ostensibly related to your circumstances. The DR plans the LLM competes with are literally just copy-pasted.
I’ve used LLMs to generate text for a soc2 audit about our processes. For one requirement, it took me 20 minutes to produce 8 pages that otherwise would have taken hours at least. This wasn’t misrepresentation or anything - more like here’s the requirement, here’s a summary of what we do, describe specifically how that meets the requirement. Outside of coding, it’s a perfect application for the tech.
Agreed, I'm using it as a technical author, if you set your system prompt up well, you get consistent style and accurate output which would take me hours of manual work.
The problem is, it's not necessarily the job of the person who is tasked with doing it.
These are the kind of things that fall between the gaps in smaller companies and there's no expert to build this disaster recovery plan because there is no risk or compliance department.
It falls into the lap of whomever is dealing with the audits or whoever has a reputation for getting things done and unblocking people.
The real management failure is taking a problem and assigning it to someone who has no interest in solving it. There's plenty of people interested in working on these kinds of things. Why waste this guy's time.
People will down vote you, but when I worked for a mega corp and a hurricane knocked out our office, they had a plan, and it worked: they set up in a hotel, and continued operations, people stayed employed, people got the products (which were largely essentials) in their supermarkets, etc.
This is not like, super trivial. People needed to figure out power, hardware, networking, vpn, etc.. etc.. etc.. and staffing. A lot of that had absolutely nothing to do with IT, but some of it did.
I've been using chatgpt for career coaching and improving my visibility at work. It has been surprisingly helpful. A million times more helpful than my own manager.
I really don't understand what the point of EMs is.
It's really just monitoring you. A designated technical lead on your team can be very helpful, but once someone moves off the dev team but also isn't able to make any decisions from the product/business side they just become dead weight. I have the same issue with my EM trying to get coaching and visibility and have gone to LLMs just like you because manager is clueless.
My skip manager started asking for weekly status reports so my manager started pulling Jira reports and feeding them to ChatGPT. It turns out that my skip manager was using Copilot to summarize those reports into basically what you could get directly from Jira.
I've seen this happen in emails too:
"Recent outage was due to a retry loop for the Foo API exceeding rate limits. We're implementing a backoff algo"
Sender, via ChatGPT:
Hi,
I wanted to provide more context regarding the recent outage.
The issue was triggered by a retry loop in the Foo API integration. When the API began returning errors, our system initiated repeated retry attempts without sufficient delay, which quickly exceeded the rate limits imposed by the API provider. As a result, requests were throttled, leading to degraded service availability.
To address this, we are implementing an exponential backoff algorithm with jitter. This approach will ensure that retries are spaced out appropriately and reduce the likelihood of breaching rate limits in the future. We are also reviewing our monitoring and alerting thresholds to detect similar patterns earlier.
We’ll continue to monitor the system closely and share further updates as improvements are rolled out.
Best regards,
Receiver, via ChatGPT:
"The outage was caused by excessive retry attempts to the Foo API, which triggered rate limiting and degraded service. To prevent recurrence, exponential backoff with jitter is being implemented"
Would be a nice fine tuning target: training to close that loop.
Recycling a prediction from a year ago:
> Ultimately a lot of this generative tech stuff is just counterfeiting extra signals people were using to try to guess at interest, attentiveness, intelligence, etc.
> So yeah, as those indicators become debased, maybe we'll go back to sending something [...] all boiled down to bullet points.
This is how society collapses and OpenAI wins.
It's already started, with AI resume spamming on one side and the associated AI assisted screening on the other. "Deep research" generated funding requests on one side and associated AI generated funding request summarization on the other.
what are the keywords to use to search for ai resume spam tools?
Unironically, "ai resume spam tools" seems to work well enough.
Even more unironically, if you run the search through Google, you'll get a nice AI summary at the top as well.
O tempora, o mores.
Also see Commandment 7 - https://muldoon.cloud/2023/10/29/ai-commandments.html
This is the kind of thing my manager and skip manager are up to. Most of our jira tickets now are written as a random jumble of ideas, fed into internal proprietary LLM and then turned into jira description and acceptance criteria. Totally pointless
Funny how encoder-decoder architectures have led to decoder-encoder behaviors.
Wow. Full circle
Plans are useless but planning is essential
~ Lots of people
This guy is quite possibly going to end up looking stupid when something goes wrong and it turns out he lied about having thought about it. I hope he is as clever as he thinks he is at anticipating what will go wrong in the future. Fires and whatnot do happen. Even AWS us-east-1 has experienced outages.
The author jokes, but almost everybody's DR plan (at least, the DR plans motivated by regs like SOC2 --- which I believe are most DR plans) are worse than what you'd get from an LLM. An LLM can at least take some input and craft something ostensibly related to your circumstances. The DR plans the LLM competes with are literally just copy-pasted.
I’ve used LLMs to generate text for a soc2 audit about our processes. For one requirement, it took me 20 minutes to produce 8 pages that otherwise would have taken hours at least. This wasn’t misrepresentation or anything - more like here’s the requirement, here’s a summary of what we do, describe specifically how that meets the requirement. Outside of coding, it’s a perfect application for the tech.
Agreed, I'm using it as a technical author, if you set your system prompt up well, you get consistent style and accurate output which would take me hours of manual work.
They're like backups, in that they don't actually exist unless you've tested them.
A copy-pasted plan is actually fine, as long as whoever's responsible for following it actually can follow it and it works.
A plan that nobody's even looked at much less tried to follow isn't fine. Even if it's word-for-word identical to one that is fine.
Or you could just do your fucking job and create a real disaster recovery plan.
The problem is, it's not necessarily the job of the person who is tasked with doing it.
These are the kind of things that fall between the gaps in smaller companies and there's no expert to build this disaster recovery plan because there is no risk or compliance department.
It falls into the lap of whomever is dealing with the audits or whoever has a reputation for getting things done and unblocking people.
The real management failure is taking a problem and assigning it to someone who has no interest in solving it. There's plenty of people interested in working on these kinds of things. Why waste this guy's time.
People will down vote you, but when I worked for a mega corp and a hurricane knocked out our office, they had a plan, and it worked: they set up in a hotel, and continued operations, people stayed employed, people got the products (which were largely essentials) in their supermarkets, etc.
This is not like, super trivial. People needed to figure out power, hardware, networking, vpn, etc.. etc.. etc.. and staffing. A lot of that had absolutely nothing to do with IT, but some of it did.
was someone keeping this plan up to date from it going obsolete?
Given the plan worked at a megacorp scale and complexity, the answer is assuredly a yes.
middle manager spotted ^
For new companies, how about "management as a service", featuring AI+MCP? No need for human managers.
I've been using chatgpt for career coaching and improving my visibility at work. It has been surprisingly helpful. A million times more helpful than my own manager.
I really don't understand what the point of EMs is.
It's really just monitoring you. A designated technical lead on your team can be very helpful, but once someone moves off the dev team but also isn't able to make any decisions from the product/business side they just become dead weight. I have the same issue with my EM trying to get coaching and visibility and have gone to LLMs just like you because manager is clueless.