Ramp Hits 30% Agent-Authored PRs as Ralph Wiggum Ecosystem Explodes with New CLIs
Daily Wrap-Up
The most striking signal from today's feed isn't any single post, it's the convergence of two separate stories into one conclusion: agent orchestration has crossed from "interesting experiment" into "production infrastructure." Ramp's @eglyman casually dropped that their internal agent, Inspect, now accounts for 30% of merged PRs in core repos, with people from "essentially every job function" submitting code. That's not a demo. That's a fundamental shift in how a real company ships software. And it happened without mandating adoption, which might be the most telling detail of all.
On the open-source side, the Ralph Wiggum pattern (autonomous agent loops with structured task management) spawned two independent implementations in a single day. @iannuttall built a CLI that works across Claude Code, Codex, and Droid, while @theplgeek shipped ralph-tui with plugin architecture and e2e observability. When multiple developers independently build tooling around the same pattern, you're watching a de facto standard emerge. The pattern is clear: PRD creation, plan generation, autonomous execution, structured tracking. Every implementation converges on these same steps because that's what actually works when you let agents cook.
The other thread worth tracking is how aggressively AI coding tools are leaking out of engineering. PMs at Fortune 500 companies are running Cursor agents against call transcripts to extract insights and generate PRDs. Non-technical operators are finding Claude Code "massive" for their workflows. The boundary between "who codes" and "who doesn't" is dissolving faster than most org charts can adapt. The most practical takeaway for developers: if you're building internal tools or workflows, design them to be agent-friendly from the start. Structure your data in repos, write clear rules files, and treat your processes as code. The teams doing this now are the ones where agents are already shipping 30% of PRs.
Quick Hits
- @steveruizok teased a "$100B app" with zero additional context. The ratio of confidence to detail is impressive.
- @steipete released bird 0.7.0, a fast X CLI for reading tweets, now with home timeline support, trending, and user-tweets. Another entry in the "rebuild social media as a CLI" genre.
- @doodlestein pointed followers to a new project that replaces their previous work, with a link but no further description.
Claude Code Expands Beyond the Terminal
Claude Code's reach continues to stretch well past its origins as a developer tool. Today's posts paint a picture of a tool finding new audiences and new use cases faster than Anthropic probably anticipated. @boringmarketer declared that "Claude Code for non technical work is massive," a sentiment that would have been confusing six months ago but now reflects a real trend of operators, PMs, and other non-engineers finding leverage in terminal-based AI workflows.
The technical users are pushing boundaries too. @DanielMiessler highlighted what he sees as the most vulnerable category of software: "kind of mid in quality, highly niche use-cases" where Claude Code can simply reverse engineer the entire product. His observation that these were previously "winner takes all" markets protected by specialized formats and protocols hits hard. The moat for mid-tier niche software was always complexity and obscurity, and that moat evaporates when an AI can just read the protocol and reimplement it.
"This is the genre of software that's in the most danger: Kind of mid in quality. Highly niche use-cases. It's been winner takes all for the space in the past. Often involved special formats or protocols. And now Claude Code can just reverse engineer it." - @DanielMiessler
On the efficiency front, @parcadei highlighted the token economics problem that plagues every AI coding session: paying 320k tokens to read a file when a tldr approach gets you the same information for 400 tokens. It's a small observation but it points to a maturing ecosystem where optimization matters. And @eyad_khrais dropped a "level 2" Claude Code tutorial, suggesting the learning curve has enough depth now to warrant tiered educational content. These tools aren't just being adopted; they're developing their own pedagogical ecosystems.
Agent Orchestration Goes Mainstream
Today might be remembered as the day autonomous coding agents stopped being a curiosity and started being infrastructure. Two independent Ralph Wiggum implementations shipped within hours of each other, and Ramp revealed numbers that should make every engineering org pay attention.
@eglyman's thread about Ramp's Inspect agent was the headliner. The system translates English requests into code, then observes reality through tests, telemetry, feature flags, and visual checks for UI work. The key insight isn't the generation; it's the feedback loop.
"One useful way to think about agents: they're control systems. Generating output is easy. Feedback is everything." - @eglyman
The numbers back it up: 30% of merged PRs authored by the agent, with adoption spreading organically across job functions. @thdxr reacted with visible shock: "it's insane how much ramp built, it's like our entire roadmap and they just went and built it for their team." When your internal tool matches an entire startup's roadmap, you've either overbuilt or you've found something real. Given the 30% PR stat, it's the latter.
Meanwhile, the open-source Ralph Wiggum ecosystem is fragmenting in the best possible way. @iannuttall built a CLI that abstracts across Claude Code, Codex, and Droid, with PRD generation and a simple ralph build command to kick off autonomous work. @theplgeek went deeper with ralph-tui, adding plugin architecture for both agents and task trackers, interactive PRD creation, dependency-aware task ordering, and e2e observability. The fact that @theplgeek built subsequent iterations of ralph-tui using ralph-tui itself is the kind of recursive validation that actually means something.
"ralph-tui is cooking. All-in-one ralph engine with e2e observability. Extensible by design. Plugin agents, plugin trackers, built-in interactive PRD creator, auto PRD conversion, customisable prompts, understands task dependencies and actionability." - @theplgeek
The convergence across all these implementations (Ramp's proprietary system, @iannuttall's CLI, @theplgeek's TUI) on the same core pattern of structured planning followed by autonomous execution with observability suggests this isn't just a trend. It's the architecture that works.
The Non-Technical Takeover
The line between "technical" and "non-technical" roles continues to blur, and today's posts suggest it might disappear entirely. @rohanvarma described a Fortune 500 PM setup that would have been unthinkable two years ago: a shared GitHub repository where customer call transcripts are committed directly, Cursor agents extract insights into dedicated directories, and PRDs are generated with a durable audit trail that other agents can reference.
"PMing in code treats product work as an evolving, inspectable system rather than a collection of docs and meetings." - @rohanvarma
This framing matters. It's not that PMs are "learning to code." They're applying version control, automation, and structured data principles to product work. The repository becomes the source of truth, not a Confluence page that nobody updates. The agents handle extraction and synthesis, not the humans. And because it's all in code, it's inspectable, diffable, and reproducible. This is what happens when the tools become accessible enough that the workflow benefits outweigh the learning curve. Combined with @boringmarketer's observation about Claude Code for non-technical work, a pattern emerges: the tools aren't dumbing down, the barrier to entry is just dropping fast enough that domain expertise matters more than coding ability.
The Developer Identity Question
Two posts today grappled with what it means to be a developer in 2026, arriving at similar conclusions from very different starting points. @ujjwalscript told the story of a 10-year veteran being outpaced by a junior who "knew how to talk to the Agents" and orchestrated three AI workers to ship a feature in 4 hours instead of 3 days. But the punchline wasn't doom. It was the PR review.
"It was fast. It was functional. But it was... fragile. It lacked architectural vision. It had security holes that only someone who has been 'burned' would see. It was a house built on sand." - @ujjwalscript
The conclusion that seniors should "compete on wisdom" rather than speed is comforting, and there's truth in it, but it undersells how quickly agents are improving at exactly those architectural and security concerns. The window where "judgment" is the differentiator is real but likely narrower than people hope.
@DaveShapi took a much more radical position, essentially advising people to prepare for permanent unemployment. His framework of remaining job categories (attention economy, experience economy, authenticity economy, meaning economy) reads like a post-labor manifesto. Whether you find it prescient or alarmist probably depends on your timeline assumptions, but the practical advice about saving aggressively and lowering cost of living is sound regardless of how AI plays out.
The Local AI Conviction
@TheAhmadOsman showed up twice today with a consistent message: run it yourself. He posted a setup running Claude Code with local models served by vLLM on 4x RTX 3090s, with GLM-4.5 Air doing the actual generation. In a separate post, he made the broader prediction: "opensource AI will win, AGI will run local, not on someone else's servers."
The practical reality is more nuanced than the evangelism suggests. Running Claude Code against local models works, but the quality gap between local and frontier models remains significant for complex reasoning tasks. Still, the economics are real. If you have the hardware, inference costs drop to electricity, and you get privacy and availability guarantees that no API can match. For teams doing high-volume, medium-complexity coding tasks, the local inference path is becoming genuinely viable rather than just ideologically appealing.
Source Posts
d
you pay 320k tokens to read a file
he pays 400 for the same file
tldr wins again https://t.co/S4XTIMEReP
S
$100B app
A
running Claude Code w/ local models
on my own GPUs at home
> vLLM serving GLM-4.5 Air
> on 4x RTX 3090s
> nvtop showing live GPU load
> Claude Code generating code + docs
> end-to-end on my AI cluster
this is what local AI actually looks like
Buy a GPU
https://t.co/WZkjjUtMoi
T
Claude Code for non technical work is massive
R
PMs aren't just using Cursor to write code. They are using Cursor to PM in code.
I spoke with a PM at a Fortune 500 company who shared their setup:
- A GitHub repository for all PMs
- Customer call transcripts checked directly into the repo
- Cursor agents extract insights from those transcripts and write them to a dedicated insights directory
- PRDs are generated into a separate folder, creating a durable record of product decisions that agents can reference later
- A robust set of Cursor Rules to guide agents through brainstorming, synthesis, and feedback workflows
PMing in code treats product work as an evolving, inspectable system rather than a collection of docs and meetings.
If you’ve discovered any interesting PM workflows with Cursor, I’d love to hear them!
I
i built a ralph cli from everything i learned from the repos and posts of @GeoffreyHuntley @ryancarson @ClaytonFarr @agrimsingh 🫡
- works with codex, claude, droid
- creates a prd for you
- turns prd into a plan
- run `ralph build` to cook
wip repo: https://t.co/LYBiYYL2NB https://t.co/Vac7rIkQJS
E
The claude code tutorial level 2
U
I’ve been a developer for 10 years.
I’ve mastered languages. I’ve optimized databases. I’ve built systems that handle millions of requests.
But last week, a Junior dev outperformed me.
He didn’t know how to write a complex program. He couldn’t explain the difference between a proper monoloth and a microservice. He didn't even know how the code worked in some parts.
But he knew how to talk to the Agents.
He orchestrated three AI workers. One for the frontend. One for the backend logic. One for the unit tests.
In 4 hours, he pushed a feature that would have taken me 3 days.
I felt a cold shiver. "Is this it?" I thought. "Am I finally the legacy hardware?"
But then I looked at his PR. It was fast. It was functional. But it was… fragile.
It lacked architectural vision. It had security holes that only someone who has been "burned" would see. It was a house built on sand.
That’s when I realized the truth about 2026.
The "Senior" title isn't about how fast you type anymore. It's about how well you judge.
We are moving from being "builders" to being "architects." From "coders" to "composers."
If you’re a veteran feeling left behind by AI: Don’t compete on speed. Compete on wisdom.
The machine can write the notes. Only you can write the symphony.
E
One useful way to think about agents: they’re control systems. Generating output is easy. Feedback is everything.
At Ramp we built a background coding agent, Inspect, that can actually translate requests in English into code, and then observe reality: tests, telemetry, and feature flags — plus visual checks for UI work (screenshots/live previews). It doesn’t just propose diffs; it iterates until the evidence says the change is correct.
Two consequences surprised me:
1. Cheap, parallel sessions change behavior. When an agent runs in a real sandboxed dev environment (not your laptop), you stop babysitting and start running more iterations.
2. Multi-client + multiplayer matters more than people think. If it shows up in the places work already happens (PRs, Slack, web, VS Code) and you can hand a session to a teammate, it becomes shared infrastructure, not a novelty.
We’re now at ~30% of merged PRs in our core repos authored by Inspect, without mandating it. People from essentially every job function, not just engineering, submitted code last week. Wild times.
ᴅ
Holy crap.
This is the genre of software that's in the most danger:
- Kind of mid in quality
- Highly niche use-cases
- It's been winner takes all for the space in the past
- Often involved special formats or protocols
And now Claude Code can just reverse engineer it. 🤯
A
calling it now, bookmark this for later
- opensource AI will win
- AGI will run local, not on someone else’s servers
- the real ones are learning how it all works
> be early
> Buy a GPU
> get ur hands dirty
> learn how it works
> you’ll thank yourself later
it’s gonna be great
J
@doesdatmaksense This new project of mine replaces that and is much better:
https://t.co/r37HLNCANo
B
ralph-tui is cooking. All-in-one ralph engine with e2e observability
- extensible by design
- plugin agents (ships with cc and @opencode plugins)
- plugin trackers (ships with json, beads, and beads-bv plugins)
- built in interactive prd creator (leverages skills)
- auto prd conversion to selected tracker format
- customisable prompts
- understands task dependencies and actionability
- quickstart
Overkill? Perhaps.
Useful? Absolutely.
A blast to use? Hell yes!
Let's go #ralphwiggum
Aiming to publish later today
@mattpocockuk @ryancarson @Steve_Yegge @doodlestein @GeoffreyHuntley
PS: Initial iteration built with ralph scripts. Subsequent iterations built with ralph-tui
D
High-agency moves to prepare for AI job loss
1) Change where you live
Imagine you learned you were never going to work again (at least not a corporate job). Where would you want to live?
Ask yourself these four questions to figure it out:
- Cultural and values alignment: where do I fit in?
- Temperature and climate: where is the weather best for me?
- Lifestyle affordances and pace of life: live fast or slow? Loud or quiet?
- Cost of living: Can I lower my COL?
2) Make good investments
Save money. Put it away. However much you make, save.
Always live below your means. If you make $80k, act like you make $60k, and save the difference. If you make $200k, act like you make $120k and save the difference.
The classical assets are
- Stocks
- Bonds
- Rental properties
I personally prefer ETFs for their accessibility and management. I don't have to worry about them. Just set and forget. Compounding returns, dividends.
My father-in-law prefers rental properties.
The rest of the details are between you and your financial advisor
3) Remaining jobs
There are a few categories of jobs that will stick around. Work towards getting one, if you want.
- Attention Economy: Become a content creator. Just be warned, this is as much about luck as it is hard work. The attention economy is "winners take most" and that's just a fact. Anyone who says otherwise is selling something.
- Experience Economy: Selling real life experiences, from bartending to massages to tours. Grounded, in-person, concrete experiences. Humans will prefer human presence forever (or at least until robots are indistinguishable!)
- Authenticity Economy: This is about trust, reputation, and "skin in the game" what some people call the "transformation economy." This includes coaches, celebrities, and so on. Stake your output to your name, your face, your voice, and your reputation.
- Meaning Economy: This boils down to priests and philosophers. Whatever your medium is (books, internet, in person) what you're doing is helping people make sense of the world they live in, and their life.
Beyond that, there will always be room for some entrepreneurs.
But things like KVM jobs, GONE.
Most knowledge work? GONE.
Low skilled labor? GONE. (replaced by robots)
4) Mission and purpose
Without work, you're going to need to reinvent your sense of purpose. Your raison d'être or your ikigai. What gets you out of bed in the morning? There are a few categories:
- Intellectual goals: do you want to get good at chess, solve big problems, and be known for your brain?
- Social goals: do you want to be recognized, famous, or influence culture?
- Dominance goals: do you want to be strong and sexy? Chase body count and glamor?
You might also want the simple life. Stick to your hobbies, your family, and your local community.
Whatever it is, you'll need to be entirely honest with yourself. This is called radical candor. Who are you really? What do you really want out of life?
---
Star working on all these today. Ideally yesterday.
You won't regret it.
P
🐦bird 0.7.0: fast X CLI for reading tweets: now with home timeline, news/trending, user-tweets and plenty fixes.
Thanks @albfresco @odysseus0z @gakonst and all other contributors who didn't list their Twitter. 😊 https://t.co/O4HZcOYt10
d
it's insane how much ramp built - it's like our entire roadmap and they just went and built it for their team