Oh yeah? If prompting is such damn cool hard thing, why can't I ask my AI slave to do all this prompting mumbo jumbo for me?
Superpowers: How I'm using coding agents in October 2025
191–200 of 242 posts
Re: Superpowers: How I'm using coding agents in October 2025
#192Earlier quoted context omitted.
Obviously if you instruct the autocomplete engine to fill in questions it will. That's not the point. The LLM has no model of the problem it is trying to solve, nor does it attempt to understand the problem better. It is merely regurgitating. This can be extremely useful. But it is very limiting when it comes to using as an agent to write code.
You can work with the LLM to write down a model for the code (aka a design document) that it can then repeatedly ingest into the context before writing new code. That what “plan mode” is for. The technique of maintaining a design document and a plan/progress document that get updated after each change seems to make a big difference in keeping the LLM on track. (Which makes sense…exactly the same thing works for human…
> that it can then repeatedly ingest into the context
1. Context isn't infinite
2. Both Claude and OpenAI get increasingly dumb after 30-50% of context had been filled
Re: Superpowers: How I'm using coding agents in October 2025
#193Is it possible to set up this kind of workflow with the plug in that comes bundled with vs code, given that you have an enterprise github copilot account that includes Claude?
Re: Superpowers: How I'm using coding agents in October 2025
#194This style of prompting, where you set up a dire scenario in order to try to evoke some "emotional" response from the agent, is already dated. At some point, putting words like IMPORTANT in all uppercase had some measurable impact, but at the present time, models just follow instructions. Save yourself the experience of having to write and maintain prompts like this.
Also the persuasion paper he links isn't at all about what he's talking about. That paper is about using persuasion prompts to overcome trained in "safety" refusals, not to improve prompt conformance.
Co-Author of the paper here. We don't know exactly why modern llms don't want to call you a jerk, or for that matter why persuasive techniques convince them otherwise. it's not a hard line like many of the guardrails. That said, I talked to Jesse about this, and I strongly suspect the same techniques will work for prompt conformance when the topic is something other than name calling.
Re: Superpowers: How I'm using coding agents in October 2025
#195> It made sense to me that the persuasion principles I learned in Robert Cialdini's Influence would work when applied to LLMs. And I was pleased that they did. No, no. Stop. What is this? What're we doing here? This goes past developping with AI into something completely different. Just because AI coding is a radical shift doesn't mean everything has changed. There needs to be some semblance of structure and design.…
> what we're getting is straight up voodoo nonsense Maybe not in this case. For the AI to create a solution, it has to come up with a vector for your intention and goals. It makes some sense for an AI trained on human persuasion materials (basically, everything has a rhetorical aspect) to also track human persuasion features for intentions. However, results will vary. Just as people trying to deploy rhetorical techni…
> It makes some sense for an AI trained on human persuasion
Why?
> However, results will vary.
Like in voodoo?
I'm sorry to be dismissive, but your comment is entirely dismissing the point it's replying to, without any explanation as to why it's wrong. "You are holding it wrong" is not a cogent (or respectful) response to "we need to understand how our tools work to do engineering".
Re: Superpowers: How I'm using coding agents in October 2025
#196Earlier quoted context omitted.
LLMs will also happily put time estimates on work packages that are based on ore-LLM turn around times. "Phase 2 will take about one week" No, Claude, it won't, because you you and I will bang this thing out in a few hours.
"Refrain from including estimated task completion times." has been in my ~/.claude/CLAUDE.md for a while. It helps.
Do such instructions take up a tiny bit more attention/context from LLMs, and consequentially is it better to leave it off and just ignore such output?
Re: Superpowers: How I'm using coding agents in October 2025
#197documents like https://github.com/obra/superpowers/blob/main/skills/testing... are very confusing to read as a human. "skills" in this project generally don't seem to follow set format and just look like what you would get when prompting an LLM to "write a markdown doc that step by step describes how to do X" (which is what actually happened according to the blog post). idk, but if you already assume that the LLM kno…
Everything is just context, of course. Every time I see a blog post on "the nine types of agentic memory" or some such I have a similar reaction.
I would say that systems like this are about getting the agent to correctly choose the precisely correct context snippet for the exact subtask it's doing at a given point within a larger workflow. Obviously you could also do that manually, but that doesn't scale to running many agents in parallel, or running automomously for longer durations.
Re: Superpowers: How I'm using coding agents in October 2025
#198Is it possible to set up this kind of workflow with the plug in that comes bundled with vs code, given that you have an enterprise github copilot account that includes Claude?
Subagents are a critical feature that GH Copilot still lacks. They allow your main agent to use another agent as a tool, meaning the main agent's context doesn't get nearly as polluted. Good read on the benefits of this pattern: https://jxnl.co/writing/2025/08/29/context-engineering-slash...
Re: Superpowers: How I'm using coding agents in October 2025
#199A big issue working with code agents is what I call context-recall: restoring context when working on a new feature or fix, that builds on recent work.
Meaning, the previous work may have involved multiple CLI sessions, summaries dumped to various markdown files like documentation files, plan files, issue files, PR-descriptions etc. Then when starting new work with a code agent you have to hunt down all of this scattered context from various md files and session logs to fill in background for the code-agent about what was recently done.
I see many workflows that help with working on a fresh feature or fix, but nothing that addresses context-recall. But maybe the OP workflow or others do that, I haven’t dug too deep into them.
Re: Superpowers: How I'm using coding agents in October 2025
#200Earlier quoted context omitted.
What you're getting at is the heart of the problem with the LLM hype train though, isn't it? "We should have rigorous evaluations of whether or not [thing] works." seems like an incredibly obvious thought. But in the realm of LLM-enabled use cases they're also expensive . You'd need to recruit dozens, perhaps even hundreds of developers to do this, with extensive observation and rating of the results. So rather than…
The UK government ran a study with thousands of developers quite recently: https://www.gov.uk/government/publications/ai-coding-assista...
I don't necessarily think the conclusions are wrong, but this relies entirely on self-reported survey results to measure productivity gains. That's too easy to poke holes in, and I think studies like this are unlikely to convince real skeptics in the near term.