Paul Geraghty

11.5K posts

Paul Geraghty banner
Paul Geraghty

Paul Geraghty

@paulgeraghty

A very dull fellow.

France, rural, very rural Katılım Ocak 2008
116 Takip Edilen300 Takipçiler
Paul Geraghty
Paul Geraghty@paulgeraghty·
@radbackwards PTT more commonly known as "push to talk", might get ya some more results.
English
0
0
0
36
dar
dar@radbackwards·
I always feel like there’s a gun to my head when I’m talking to LLMs in voice mode. I can’t take a breath or think for a second without it prematurely answering. A walkie-talkie style hold-to-speak would solve this.
English
26
0
158
5.4K
Paul Geraghty
Paul Geraghty@paulgeraghty·
Radical thought to improve soccer games. If its 0:0 at half time, use two footballs on the pitch. Giddy-up!
English
0
0
0
8
Michael
Michael@michaelmcguk·
Yo @MichaelBuble, perfume?? I think you'd be better selling buble bath !
Michael tweet media
English
1
0
1
41
Paul Geraghty
Paul Geraghty@paulgeraghty·
Waiting for a new company to emerge with an Agent and Skill combo that gradually dulls your codebase, adds broken links, out of date libs, todo's never completed. Some spaces become tabs. Orphan methods and classes are introduced willy-nilly. Yeah, that's right, Entropic.
English
0
0
0
15
Paul Geraghty retweetledi
Damian Edwards
Damian Edwards@DamianEdwards·
Copilot CLI plugin/extension to show real-time session cost, including currency conversion, PRU vs. UBB, & "what-if" subscription type support? It's here 😁 github.com/DamianEdwards/… *these are best-effort cost estimates, 100% vibe-coded, feedback welcome
Damian Edwards tweet media
English
6
16
89
11K
Paul Geraghty
Paul Geraghty@paulgeraghty·
Back from #GOSIM #GOSIMParis 2026, some interesting talks, met friends, but my big takeaway can be summarised by the recurring topic, in almost every talk/presentation/conf the word TRUST, the lack of it, and what we can do about it, or not.
English
0
0
0
26
Paul Geraghty
Paul Geraghty@paulgeraghty·
uncanny, but I remember 'skimming' code looking for a method, because I remembered its shape (actual shape on the page), not its name, nor really it actual job, probably just a jammy attempt to find some duplicate reasoning to refactor out.
Uncle Bob Martin@unclebobmartin

@chrisbward To some extent, yes. But far more important than the shape of the code is the interconnectivity of the code and the organization of the code. And that is not encoded in the shape of the code.

English
1
1
24
6.3K
Paul Geraghty retweetledi
Uncle Bob Martin
Uncle Bob Martin@unclebobmartin·
Morning bathrobe rant: disengage from the syntax.
English
50
51
635
39.4K
Zara Zhang
Zara Zhang@zarazhangrui·
I realized a lot of people treat coding agents as their employee, whereas I actually treat it as my cofounder I don't just give orders. I present problems, describe the situation, and ask for their opinion
English
36
11
156
10.3K
Paul Geraghty
Paul Geraghty@paulgeraghty·
@zarazhangrui just because you ask their opinion, does not mean you have to accept it. Thats just good management. (er, if you do accept the plan does that mean the agent/llm is more motivated to make sure it works out well, like a real employee would?)
English
0
0
0
136
Paul Geraghty
Paul Geraghty@paulgeraghty·
Spot on. Good man management and leadership techniques, patterns and that means, gasp, ANTI-PATTERNS! Oh boy, gonna start collecting them now. I bet this arena is bursting with them.
Alex@alexgreensh

@zarazhangrui That's how good managers actually treat their employees TBH... Giving orders and micromanaging is toxic for both humans and agents. With humans it's about Culture. With agents it's all about Context.

English
0
0
0
15
Paul Geraghty
Paul Geraghty@paulgeraghty·
This. Also, watch bloomergs' Wall Street Week, 15m in to 30m. Guy from Google saying much the same, but not as pointedly. I thought it was just me.
Sukh Sroay@sukh_saroy

A new study just blew up the entire "vibe coding" movement. Researchers from UC San Diego and Cornell tracked 112 experienced software developers using AI agents in their actual jobs. The finding is the opposite of every viral demo on your timeline. Professional developers don't vibe code. They control. Here's what they actually found. The researchers ran two studies. 13 developers were observed live as they coded with agents in real production work. 99 more answered a deep qualitative survey. Every participant had at least 3 years of professional experience. Some had 25. The viral pitch of agentic coding goes like this. Hand the agent a vague prompt. Don't read the diff. Forget the code even exists. Trust the vibes. Andrej Karpathy coined the term. Tens of thousands of developers on X claim to run "dozens of agents at once" building entire production systems hands-off. The data says almost nobody serious actually works that way. Here is what experienced developers do instead. → They plan before they prompt. They write out the architecture, the constraints, and the edge cases first, then hand the agent a tightly scoped task. → They review every diff. Not because they're paranoid. Because they've seen what happens when you don't. → They constrain the agent's blast radius. Small, well-defined tasks only. The moment a problem touches multiple systems or has unclear requirements, they take over. → They treat the agent like a fast junior dev that needs supervision, not a senior engineer that can be trusted alone. The researchers also found something darker buried in the data. A separate randomized trial they cite showed that experienced open source maintainers were 19% slower when allowed to use AI. A different agentic system deployed in a real issue tracker had only 8% of its invocations result in a merged pull request. 92% failure rate in production. 19% productivity drop for senior devs. The viral demos lied to you. The paper's biggest insight is in one sentence: experienced developers feel positive about AI agents only when they remain in control. The moment they let go, quality collapses, and they know it. This matches what every serious shop has quietly figured out. The developers shipping the most with AI right now aren't the ones vibing. They're the ones with the strictest review processes, the tightest task scoping, and the clearest mental model of what the agent can and cannot do. Vibe coding makes for great Twitter videos. It does not make great software. The next time someone tells you they let Claude build their entire SaaS in a weekend, ask them how much of that code they've actually read. The honest answer separates real engineers from the demo crowd.

English
0
0
0
21
Paul Geraghty retweetledi
Sukh Sroay
Sukh Sroay@sukh_saroy·
A new study just blew up the entire "vibe coding" movement. Researchers from UC San Diego and Cornell tracked 112 experienced software developers using AI agents in their actual jobs. The finding is the opposite of every viral demo on your timeline. Professional developers don't vibe code. They control. Here's what they actually found. The researchers ran two studies. 13 developers were observed live as they coded with agents in real production work. 99 more answered a deep qualitative survey. Every participant had at least 3 years of professional experience. Some had 25. The viral pitch of agentic coding goes like this. Hand the agent a vague prompt. Don't read the diff. Forget the code even exists. Trust the vibes. Andrej Karpathy coined the term. Tens of thousands of developers on X claim to run "dozens of agents at once" building entire production systems hands-off. The data says almost nobody serious actually works that way. Here is what experienced developers do instead. → They plan before they prompt. They write out the architecture, the constraints, and the edge cases first, then hand the agent a tightly scoped task. → They review every diff. Not because they're paranoid. Because they've seen what happens when you don't. → They constrain the agent's blast radius. Small, well-defined tasks only. The moment a problem touches multiple systems or has unclear requirements, they take over. → They treat the agent like a fast junior dev that needs supervision, not a senior engineer that can be trusted alone. The researchers also found something darker buried in the data. A separate randomized trial they cite showed that experienced open source maintainers were 19% slower when allowed to use AI. A different agentic system deployed in a real issue tracker had only 8% of its invocations result in a merged pull request. 92% failure rate in production. 19% productivity drop for senior devs. The viral demos lied to you. The paper's biggest insight is in one sentence: experienced developers feel positive about AI agents only when they remain in control. The moment they let go, quality collapses, and they know it. This matches what every serious shop has quietly figured out. The developers shipping the most with AI right now aren't the ones vibing. They're the ones with the strictest review processes, the tightest task scoping, and the clearest mental model of what the agent can and cannot do. Vibe coding makes for great Twitter videos. It does not make great software. The next time someone tells you they let Claude build their entire SaaS in a weekend, ask them how much of that code they've actually read. The honest answer separates real engineers from the demo crowd.
Sukh Sroay tweet media
English
199
335
1.7K
264.6K
Paul Geraghty retweetledi
Uncle Bob Martin
Uncle Bob Martin@unclebobmartin·
I’ve been harping on the disciplines and tools for using AI lately. I find them to be a very effective approach. But I don’t want to leave you with the impression that a few simple disciplines and tools is sufficient. As the AI’s build software, you — the software engineer — need to have a good mental model of what the AI is doing. You need to apply engineering insight to correct it when it takes a path you don’t like. You have to be an active manager in the design and architecture of the system. You have to be able to “see within“ without resorting to exhaustive code reviews. You have to form suspicions about what the AI is doing, and you have to probe and experiment to verify your suspicions.
English
54
82
823
36K
Paul Geraghty
Paul Geraghty@paulgeraghty·
for english ppl living in france. artybollocks.com "Oh, you're english, you must be an artist then ..." Usually correct.
English
0
0
0
6
Michael
Michael@michaelmcguk·
@paulgeraghty I get to fast track at any Greggs within a 30 mile radius. Truly living.
English
1
0
0
14
Paul Geraghty
Paul Geraghty@paulgeraghty·
@michaelmcguk yeah, employee of the year over at Greggs, i heard. Pride of place on the mantlepiece.
English
1
0
1
16
Michael
Michael@michaelmcguk·
@paulgeraghty It just has to be. Mad to think that's been nearly 12 years. I still get stopped in the street.
English
1
0
0
12
Michael
Michael@michaelmcguk·
@paulgeraghty Still good news on the squashed trophy hunter ! :-D
English
1
0
0
8