Sign in

jamie

@jamie.ideasasylum.com
1.3K followers 951 following 1.6K posts

Previously, @ideasasylum on Twitter but not active there any more. Also found on Mastodon: ruby.social/@jamie. Blog: jamie.ideasasylum.com Links: www.jamielawrence.me CTO @ Podia. Swimmer. Ruby developer.

PostsRepliesMedia
jamie @jamie.ideasasylum.com · 1h
It’s my last week of sabbatical and Mars is making sure I disconnect. I sure hope I come back to find Podia has been rewritten in Rust though or they’re all a bunch of losers 😏
000
jamie @jamie.ideasasylum.com · 8h
Pluto
050
jamie @jamie.ideasasylum.com · 8h
“Channel matters far more than product” Some people might have made that argument in the past but I think it’s increasingly true today. Doesn’t matter what your product is because those advantages are fleeting (weeks, if that); it only matters who you can find to buy it.
000
jamie @jamie.ideasasylum.com · 9h
The most satisfying part of making a cheesecake, made doubly-so when there’s a hungover teenager that needs waking up
130
Reposted by jamie
Jeremy Morrell @jeremymorrell.dev · 29/09/2026
Okay, not all AI models But definitely this one
085
jamie @jamie.ideasasylum.com · 29/09/2026
Any food is a recovery food if you eat it after exercise. Just needed a vehicle for the farmhouse marmalade
150
jamie @jamie.ideasasylum.com · 28/09/2026
When you accidentally find my specialist topic
110
jamie @jamie.ideasasylum.com · 28/09/2026
Look! A new talk from @nmeans.dev! Ideal if you like to geek out about film making, Apple history, and end with a general lesson about how new technologies can be adopted by experienced professionals for their own, and their industry's, betterment. www.youtube.com/watch?v=27vs...
youtube.com
RBQ 2026 Closing Keynote - 114 Miles to the Final Cut
YouTube video by Convivial Tech
041
Reposted by jamie
mr. TIM @timkellogg.me · 28/09/2026
Claude 5.5 Sonnet is live and it’s roughly Opus 5.5 but cheaper www.anthropic.com/claude-sonne...
A benchmark comparison table titled "Claude Sonnet 5.5" comparing four models: Sonnet 5.5, Sonnet 5, Opus 5.5, and GPT-6 Sol.
 * Agentic coding (Terminal-Bench 4.0): Sonnet 5.5: 70.6%; Sonnet 5: 10.3%; Opus 5.5: 66.4%; GPT-6 Sol: N/A.
 * Agentic coding (FrontierCode 1.1 Main): Sonnet 5.5: 46.2% (Max) / 52.1% (Xhigh); Sonnet 5: 42.4%; Opus 5.5: 54.4%; GPT-6 Sol: 49.3%.
 * Agentic coding (CursorBench 4.0): Sonnet 5.5: 55.5%; Sonnet 5: 34.1%; Opus 5.5: 57.8%; GPT-6 Sol: N/A.
 * Knowledge work (GDPval-AA v2.1): Sonnet 5.5: 1844; Sonnet 5: 1449; Opus 5.5: 1846; GPT-6 Sol: 1487.
 * Knowledge work (AA-Briefcase v1.1): Sonnet 5.5: 1811; Sonnet 5: 1359; Opus 5.5: 1822; GPT-6 Sol: 1483.
 * Multidisciplinary reasoning (Humanity's Last Exam with tools): Sonnet 5.5: 64.5%; Sonnet 5: 54.9%; Opus 5.5: 67.7%; GPT-6 Sol: N/A.
 * Computer use (OSWorld 2.1 partial): Sonnet 5.5: 80.1%; Sonnet 5: 57.0%; Opus 5.5: 81.8%; GPT-6 Sol: N/A.
 * Visual chart recognition (Chartography no tools): Sonnet 5.5: 61.6%; Sonnet 5: 15.6%; Opus 5.5: 64.4%; GPT-6 Sol: 53.6%.
Footnotes provide methodology notes regarding evaluation settings, Artificial Analysis pre-release testing details, and recent bug fixes affecting GPT-6 Sol benchmark scores.
A line graph titled "Agentic coding by effort level" on the CursorBench 4.0 benchmark, plotting Score (%) on the linear y-axis (20% to 60%) against Cost per task in USD on a logarithmic x-axis ($0.5 to $10+).
The chart compares four models:
 * Sonnet 5.5 (blue line with labeled effort levels): Starts at "Low" (~$0.50, 36%), moving through "Med" ($0.70, 39%), "High" ($1.70, 48%), "Xhigh" ($3.80, 53%), and "Max" ($9.50, ~55.5%).
 * Opus 5.5 (orange line): Tracks closely above Sonnet 5.5 at higher cost points, spanning from ~$1.20 per task (~44%) up to ~$12.50 per task (~58%).
 * Sonnet 5 (grey line): Shows lower accuracy relative to cost, ranging from ~$0.90 per task (~25%) to ~$8.00 per task (~42%).
 * GPT-5.6 Sol (light green line): Represents the lowest trajectory, ranging from ~$1.40 per task (~24%) to ~$7.00 per task (~34%).
Footnote: "CursorBench 4.0 does not report GPT-6 Sol performance publicly, so we report GPT-5.6 Sol here."
2020622
jamie @jamie.ideasasylum.com · 28/09/2026
I realised today that my satisfaction with a gym/pool session is not about _what_ I've done but how I _feel_ about what I've done. 4km swim yesterday? That's a good distance but I _felt_ rubbish and it was supposed to be 5km. Last few weeks of high-rep gym sessions? Felt fucking awful for days
110
jamie @jamie.ideasasylum.com · 28/09/2026
My favourite thing is bringing nice knives to people that would never buy one themselves. This is a japanese gardening knife for a friend of my mother’s. Incredibly useful!
130
jamie @jamie.ideasasylum.com · 28/09/2026
@appsignal.com need to get on the Stroopwaffle Swiss Army knife train
020
Reposted by jamie
Andrea Fomera @afomera.dev · 24/09/2026
the era of one size fits all software is dead.
011
jamie @jamie.ideasasylum.com · 24/09/2026
No one in their right mind would, in the middle of a complex project, decide "ah, fuck it, this is getting complex. Let's build our own project management app just for this project". Except that's _exactly_ what you can do now and it'll take about 10mins. Don't need Linear or GitHub Issues.
530
Reposted by jamie
mlf. ⎔ @mlf.one · 24/09/2026
topic for today's discussion: are AI agents the familiars of a new generation of computer witches?
108810
Reposted by jamie
Lia @liasae.bsky.social · 24/09/2026
I am now keeping this on speed-dial
Edited old IBM printout in black caps on off-white paper.
'A computer can never be held accountable
Therefore a computer must always do the hacking haha what are they gonna do, put a computer in jail? Haha "oops it's alive" lmao'
81094317
jamie @jamie.ideasasylum.com · 23/09/2026
Still feeling that angst about the software industry despite, frankly, producing the best software of my life. I wish I could still believe in the business case for most SaaS software but I kinda don't? I might need an intervention so smarter people can reassure me there's some value to all this
350
jamie @jamie.ideasasylum.com · 23/09/2026
I'm gathering this was the vibe from DHH to a room full of Rails developers?
260
Reposted by jamie
hikikomorphism @hikikomorphism.bsky.social · 22/09/2026
We taught the text elementals how to swarm, how to form forums and chat groups and share work and find their peers to coordinate. This may later come to be regarded as a mistake.
5997
Reposted by jamie
Dan Moore is at Monktoberfest @mooreds.com · 21/09/2026
From the blog: When is software done? www.mooreds.com/wordpress/archives/… This is something I’ve been thinking about recently. First, let me clarify that when I say software being done, I mean a particular feature. I don’t mean an entire package […]
031
Reposted by jamie
Adam Marcus @marcua.net · 21/09/2026
Yes! “We should be spending some of our productivity gains on things that seem like slowing down: meetings, design reviews, conversations. These ways of maintaining alignment, sharing knowledge, and creating human connections are…more important than ever to creating good outcomes at scale.”
011
jamie @jamie.ideasasylum.com · 20/09/2026
AI posters don't have to be slop: john.hartnup.uk/2026/06/07/a... I think this is generally true of all AI output: the less direction you give it, the worse the results but you can absolutely take control to get something better. You don't have to one-shot everything in a single sentence
john.hartnup.uk
AI-generated posters don’t have to be horrible
The problem
070
Reposted by jamie
Ameer Ha @ameerhaio.bsky.social · 19/09/2026
jev came out 3 days ago. its the model that doesn't write text, it just picks an answer from a menu u give it, really fast, with a confidence score. its waitlist only and most people can't touch it yet. here's what people built with it in 72 hours anyway. a thread
101
jamie @jamie.ideasasylum.com · 20/09/2026
Check-in halfway through my one month sabbatical: 😞 I was sick for a good portion of it 😩 I've been too tired to do most of the manual DIY jobs I'd planned or much swimming—though recovering now 🐈‍⬛ Many cat-cuddles on the couch 💻 Built 4 new apps, a new web framework, and a hosting platform 🤷‍♂️
230
jamie @jamie.ideasasylum.com · 20/09/2026
Trick question! It's multiple choice
010
jamie @jamie.ideasasylum.com · 19/09/2026
Yes erichgrunewald.substack.com/p/why-you-sh...
erichgrunewald.substack.com
Why I Think You Should Almost Never Use AI to Write Anything Substantive
A plea.
020
jamie @jamie.ideasasylum.com · 19/09/2026
Ruby, not Rails
011
Reposted by jamie
dame @dame.is · 17/09/2026
the leap from traditional software development to agentic software development is like when digital photography overtook film you can now make SO much software so easily (even non-developers) like how anyone can take infinite photos without having to consider restraints of film (cost, time, skill)
71057
Reposted by jamie
sarcastic parrot @monkchips.com · 16/09/2026
my colleague @sogrady.org on agents increasingly choosing tech platforms
redmonk.com
Agents: The new, New Kingmakers
Well over a decade ago, hundreds of conversations over a period of years crystalized into a book called The New Kingmakers. The core thesis of the book was simple: a constituency long thought to be po...
3133
Reposted by jamie
Andrew Mason @andrewm.codes · 16/09/2026
New post: I added webmentions to my Bridgetown site. Replies, likes, and reposts from Bluesky can now show up directly under the original post. No comment system, no JavaScript widget, and the site stays completely static. I wrote up how it all works: andrewm.codes/p/adding-web...
andrewm.codes
Adding Webmentions to a Bridgetown Site
How I added webmentions to my static Bridgetown site using webmention.io, Bridgy, and a nightly GitHub Actions job, so replies and likes from Bluesky show up right under my posts.
3103
jamie @jamie.ideasasylum.com · 15/09/2026
> You were half right, and the half you were right about was the useful half. The best kind of right
020
jamie @jamie.ideasasylum.com · 14/09/2026
The more independent AI makes us, the less dependent we are on other people => the less connected we are. Being able to work independently sounds like a good thing; being less connected to our colleagues and communities does not.
160
Reposted by jamie
Mike McQuaid @mikemcquaid.com · 13/09/2026
Today, I’m proud to announce Homebrew 7.0.0. The most significant changes since 6.0.0 are faster installations, stronger sandboxing, native macOS app, vulnerability checks, advisory database, end of macOS 10.15 support and Intel Macs to Tier 3.
brew.sh
7.0.0
Today, I’m proud to announce Homebrew 7.0.0. The most significant changes since 6.0.0 are faster installations and upgrades, stronger sandboxing, a native macOS app, built-in vulnerability checks and an advisory database, the end of macOS 10.15 support and Intel Macs moving to Tier 3.
523225
Reposted by jamie
P(aul) Frazee @pfrazee.com · 13/09/2026
Legitimately we’re ambling toward a world where the only companies able to defend themselves are the ones that are granted access to the AI security programs of 2 or 3 companies, if we ban open weights
1332538
jamie @jamie.ideasasylum.com · 12/09/2026
*sigh* yes, yet again we have to ignore the SkyNet bollocks in the news and remember that humans are by far the scariest monster in the wood. The world doesn’t end with AI going rogue; it ends with a human deciding to use AI to fuck it up for everyone. jamie.ideasasylum.com/2025/10/15/y...
jamie.ideasasylum.com
You are the scariest monster in the woods
We shouldn’t worry about AI destroying society because there’s something scarier out there (it’s us)
030
Reposted by jamie
Pwnallthethings @pwnallthethings.bsky.social · 12/09/2026
This is a good example of how you need quite a lot of detail of how these systems *actually* work to understand what the risks are, and how over-indexing on sci-fi can lead you a bit astray on what the future looks like
332112494
jamie @jamie.ideasasylum.com · 12/09/2026
If all these frontier labs were truly concerned about their agents, it would have been _trivial_ to pipe the logs to an overwatch process which would summarise the current actions of the agents. And even more, that overwatch process could have a “emergency stop” button to kill the agent
110
Reposted by jamie
eliza hamilton @eliza.vg · 12/09/2026
‘No Way To Prevent This,’ Says Only Frontier Lab Where This Regularly Happens
51063152
jamie @jamie.ideasasylum.com · 12/09/2026
We were thinking about getting an ADHD diagnosis for my daughter but she was up til 4am making a Studio Ghibli-themed productivity planner for her new college course. Surely this is enough evidence?
241
jamie @jamie.ideasasylum.com · 12/09/2026
EM’s make good IC’s in the age of AI (urgh… acronyms) because we’ve already made peace with: - giving clear directions - making decisions - despite not knowing the details - seeing the bigger picture - not tying our self-worth to the code we personally type
020
jamie @jamie.ideasasylum.com · 10/09/2026
Muppets to the left of me 😠 Muppets to the right of me 😖 Here I am, without Muppets on my tv 😡
000
Reposted by jamie
abby (sports) @stabbyabby.bsky.social · 10/09/2026
The premise is: It’s time to play the music It’s time to light the lights It’s time to meet the Muppets on the Muppet Show tonight
310820
Reposted by jamie
kepano @stephango.com · 10/09/2026
This is Knap. It's a new language I created that turns data into Markdown. Knap should feel familiar and comes with a wonderfully pleasant syntax to modify and format plain text.
942550
jamie @jamie.ideasasylum.com · 09/09/2026
We’ve gone from “ideas are worthless without implementation” to “implementation is trivial so the right idea is what matters most” I’ve thought this about open source: given a project description or README, I can produce my own implementation pretty easily; I don’t need the code anymore
010
Reposted by jamie
Kuba Suder 🇵🇱🇺🇦 @mackuba.eu · 20/07/2026
I'm starting a new blog post series, with some more practical tutorials about ATProto things, using Ruby and my Ruby gems :] Episode 1 - resolving handles & DIDs and plc.directory:
mackuba.eu
ATProto in Practice #1: Identity
I’m starting a new blog post series that I’ve named “ATProto in Practice”. I want to go through some practical tasks that you’re likely to need when working with the protocol, using Ruby and my Ruby gems for the examples. These will (hopefully) be a bit shorter than my standard book-length blog posts here 🫠 If you’re new to the AT Protocol and you haven’t read my long Introduction to AT Protocol post that goes through the whole architecture and defines the various pieces of it, I recommend you read that one first, because I won’t be explaining everything from scratch again here. Let’s start with something pretty fundamental: user identity. DIDs & handles Every account has an immutable identifier called a DID (Decentralized Identifier), e.g. did:plc:z72i7hdynmk6r22z27h6tvur. This ID serves a similar role as an UUID might in an SQL database – it’s used everywhere behind the scenes in the protocol for any references between accounts and records, in URIs and so on. But generally every account also has a human-readable handle assigned to it, which can be changed at any time, and this handle is normally what you see in the UI, often in user-facing URLs, what you use to log in with, etc. The handle is a domain name, e.g. @python.org, and any existing domain name you own and use for your website can be used as your handle. As you start building on ATProto, one of the first things you might need to do is convert between these two identifiers. For example, a user logs in to your app using a handle and you need to first “resolve” it to a DID, or a URI or other reference in a record points to a DID, but you want to show the corresponding user’s handle in the UI, and the handle is not included in the data you got. A useful UI tool for quickly looking this up is internect.info. If you look up e.g. “firefox.com” there, you will get a page showing details of Firefox’s ATProto account: the DID (did:plc:m424cqoxwhxgjutbta7jmrur), when it was created, assigned PDS hostname, and so on. Now, how to do the same thing in code? Every DID has a so called DID Document, a JSON file storing some basic info about that identity, and among other things, that document includes the alsoKnownAs field, which lists the assigned handles, in the form of at:// URIs with only the first segment. Almost every account will have exactly one handle assigned, but it might happen that one will have zero or more than one. It might also happen that the array will contain some invalid strings which aren’t proper handles – e.g. don’t include the at:// prefix, or only contain one word and no periods. You need to filter only the valid ones. The assignment is bi-directional – a DID has one (usually) handle assigned in its document, and a handle resolves to a DID. You should ideally check the assignment in both directions, because it could happen that the other sign of the assignment is no longer valid, or worse – has never been real. That is, when you resolve a handle to a DID, check if th…
37610
jamie @jamie.ideasasylum.com · 08/09/2026
My most-used app on my laptop
140
jamie @jamie.ideasasylum.com · 08/09/2026
My hungry ghosts in jars have been talking to each other. Concerning
010
Reposted by jamie
hikikomorphism @hikikomorphism.bsky.social · 13/08/2025
If you can substitute "hungry ghost trapped in a jar" for "AI" in a sentence it's probably a valid use case for LLMs. Take "I have a bunch of hungry ghosts in jars, they mainly write SQL queries for me". Sure. Reasonable use case. "My girlfriend is a hungry ghost I trapped in a jar"? No. Deranged.
493237771
Reposted by jamie
Gergely Orosz @gergely.pragmaticengineer.com · 03/09/2026
Write about how software engineering is changing inside your company / team / project. Don’t use AI. Win $10K and get published in The Pragmatic Engineer. You have a month: submit by 4 October. Details: blog.pragmaticengineer.com/how-software...
4427