Michael Bleigh

Waiting for a reply shaped like this on this post: "A slop button changes the game on automated posters. One click and the post is gone. It's not just cleaning up the timeline, it's changing the incentive structure." 🙃

Every social media service needs: 1. A "slop" button that is a combination user mute and report of a post as being low-quality. 2. ToS allowing discretionary banning of users for automated low-quality posts. Right now slop replies are clearly rewarded. They need to be punished.

@Noahpinion Slop means "low-effort, mimicking the fat part of the distribution". Last Jedi is neither of those things. People hate it because it tried an out-of-distribution take on a mainstream property and didn't really pull it off.

The Last Jedi was slop

@MattBruenig @TheStalwart @DKThomp @jdcmedlock I could dedicate my life to being the best in the world at bashing my head against the wall, work 18 hours a day for 10 years. Best head basher on the planet. Who else benefits from my effort? George Lucas made a thing that millions of people love, seems worth rewarding.

@TheStalwart @DKThomp @BigMeanInternet @jdcmedlock Ha. Well let's put aside whether you think writing Chewbaca is more impressive than a 2 hour marathon. Surely you think a 2-hour marathon is impressive and can see that tons of people train for running. So what would be your objection to making them billionaires through that law?

Obama was so real for this

The dark part is imagining what we'd hook up to "pain" receptors in conscious machines. Lint failure in its code gives it a little jolt, security hole puts it in the "hospital". But even things like pain and time perception are likely to be "simulated" soon. Where's the line? I'd posit that we will never be able to establish a bright line test for consciousness and recognition of artificial personhood eventually becomes a divisive civil rights issue.

People are completely missing Richard’s point here. I’d like to think it’s because they read only until the paywall. He is not saying that LLMs are conscious. Instead he’s saying: 1. LLMs are deeply capable (this is true) 2. If LLMs are not conscious (they almost certainly aren’t) then they are ‘mindless zombies’ that approach/surpass humans in capability. Based on this, he asks: If capability can be mindless or conscious, why did the most intelligent creatures on this planet evolve consciousness, even though it’s not strictly necessary? It must have some inherent value, like the ability to feel pain. This is a pretty sound argument, and I’ll agree that consciousness adds value. The unsaid follow-up question is: Will machines at some point be conscious? Is that the next step in their evolution which will make them yet more capable and valuable? Will they at one point be able to perceive pain, the movement of time, and other elements we currently associate with consciousness? They can’t now, but if we look 10-100 years in the future it would not be ludicrous to believe that machines will be not only capable, but also conscious.

@samuelcolvin Check out genkit.dev

TypeScript people - I have a question: What's the best (least bad, most trendy) Agent Framework right now? Vercel AI, Mastra, Langpain-js - any other popular options? I mean for general applications (e.g. not coding agents). Do you even use a framework, or pretend you don't need one and let the coding agent one-shot a slop micro-framework each time?

@StatisticsFTW I think it's the biggest failure of GraphQL docs and ecosystem onboarding that this isn't in the first 10 minutes of being introduced to GraphQL. It's like teaching people how to use guns without telling them safeties exist.

I think it's underappreciated that GraphQL best practices include "don't send GraphQL over the wire". Instead, it's "do a generated RPC with GraphQL responses." Which addresses 99% of the concerns that folks have when they complain about GraphQL 😅

Yes AI will be able to do things with "competent" taste because that's the fat part of the distribution. But to do something that *innovates* on design and experience you need an insight that is inherently out of distribution. That's the potential moat. Regardless, humans with discerning taste will be able to get a lot more out of AI even as AI becomes much more capable.

my hot take is that "taste is a uniquely human thing" is cope. one simple example: i've noticed that friends of mine who are not designers at all (for example, backend / infra people) were making projects with ai 2 years ago that looked obviously sloppy. now the same people are shipping things that look… pretty polished. probably good enough that a junior designer could pass it off as their own work. and when i ask them what changed -- did you learn some design fundamentals? did your 'visual taste' improve? are you better at noticing bad spacing / typography / color hues? most of the time the answer is just: the tools got better. the models got better. i can only judge this for visual design / frontend because that’s where i feel like i have some taste. but i’m pretty sure the same thing is happening in copywriting, backend architecture, research ideas, product taste, etc. yes, prompting matters. yes, having a good taste yourself still matters. but the median 'taste' of an ai model in 2023 and the median taste of an ai model in 2026 are very different. there is a waterline of taste and it is rising very fast. sloppification of the internet was a temporary phase.

Okay so it turns out Telegram is 1000x easier to set up with a bot than WhatsApp or Google Chat or SMS. I don't use Telegram generally, I didn't particularly want to start, but I might for this reason alone.

I wanted to play around with building a personal agent for my family and the process of getting working credentials to talk to a chat app is...next to impossible? Is there some secret shortcut to doing this that doesn't require me to register an LLC or something?

@rockwotj @TaskletAI But I want to hack it together myself out of bubblegum and duct tape!

@mbleigh Use @TaskletAI 😋

This was something I really wanted in Go, but now it's hard to be excited because I basically don't write code by hand anymore. A rare moment of post-AI melancholy for me.

@mjackson And, regrettably, Python.

Accurate

@stanislavfort @mooncat_is What's the false positive rate though?

Fair pushback on the needle-in-haystack stuff @mooncat_is. So I actually built our nano-analyzer as full-repo scanner & it works! Parallelism lets small models (4B) catch the FreeBSD CVE & find new kernel bugs Nano-analyzer: github.com/weareaisle/nan… x.com/stanislavfort/…

@GergelyOrosz If you need humans to review the code (still true for most production codebases at most orgs I suspect) stacked diffs are the only way to work. You spike a big change, shape it out to where you like it, then ask the agent to break it into a chain of stacked commits for review.

Better late than never, but stacked diffs feels like it was relevant pre-2024/2025 for eng teams big on using AI: before AI agents became big, and started to generate large code changes. I'd go as far to say that stacked diffs make most sense when writing a lot of code by hand

I think the most common frustrating architecture decision I run into is: - To do X (thing we want to do a lot, like define a new tool for an agent) - You need to add code to Y, Z, A, and B Features that need to scale should have vertical consolidation of their definitions.

Which of the chat platforms are adding streaming message support for better UX for agents in chat?

@ClementDelangue A bloodhound searches five square miles and discovers an errant handkerchief. A person stands and points at the handkerchief, and a child walks over and picks it up. Did the child "find" the handkerchief?

"But here is what we found when we tested: We took the specific vulnerabilities Anthropic showcases in their announcement, isolated the relevant code, and ran them through small, cheap, open-weights models. Those models recovered much of the same analysis. Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens. A 5.1B-active open model recovered the core chain of the 27-year-old OpenBSD bug." aisle.com/blog/ai-cybers…

@FutureVibesOnly @AnthropicAI @grok Kilocorn = 1K Unicorns :)

@AnthropicAI What even is this business?! Fucking triple the run rate in 3 months? Amazing. They can't IPO soon enough. A unicorn is a 1 billion pre-IPO valuation. What do we call these 1 TRILLION pre-IPO companies? Help me out @grok.

Our run-rate revenue has surpassed $30 billion, up from $9 billion at the end of 2025, as demand for Claude continues to accelerate. This partnership gives us the compute to keep pace. Read more: anthropic.com/news/google-br…

@dbreunig True. I think there's probably a way to trace many "whys" even by prompting around and exploring the history of the code in specific ways, examining commit messages (and linked PRs from commit messages), etc. Seems like an agent skill someone could put together.

@mbleigh And so often the answer is, "Oh, that's slop."

An interesting thing about the AI era is you no don't need humans to figure out "how" or "what". New to a codebase and need to trace a code path or discover relation between components? Ask an agent. Now most questions to my teammates are about rationale or judgment - the "why".