Erik Edin (@321K) header
Erik Edin (@321K) profile picture

Head of Data at Marcura. Ex founder, ex-@wise.

Stockholm, Sweden
Joined April 2009
Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)Photo by Erik Edin (@321K)
In terms of peak AI spend, I don’t think Antrophic could have timed their IPO any better. Even in the enterprise, 2026 was all about tokenmaxxing. 2027 will be about ROI.
1
107
A lot of AI agent tools are just really slow reverset ETL pipelines.
1
1
170
out of all the writing I’ve encountered about raising little kids, I think this is the piece I’m most grateful to have read
46
1,591
25
48,574
719,529
My little 5x5 workout app now has an official ChatGPT plugin.
4
211
Erik Edin retweeted
I am doing a project about elderly vibecoders If you are a vibecoder and over 25 please DM
1,140
210
238
7,701
388,475
Why is it so hard to read Claude's PR reviews? Is there a way to fix it?
1
74
My gym visits · 5×5
1
86
Prompt injecting my colleagues granola recordings
1
68
This calls the viability of the entire internet into question. If simply using any online software requires paying enormous taxes to Anthropic and OpenAI in the form of huge cyber bills, will it still be worth it to use all the software we use now?
some quick thoughts on frontier labs and cybersecurity 1) two critical vulnerabilities were discovered in php that allow for remote code execution; the first was reported on august 11 2) it seems now humans just weren’t devoting so many resources to finding vulnerabilities, and perhaps weren’t that good at it in the grand scheme of things 3) i think open-source software is going to have to change: maintaining it will increasingly become the responsibility of frontier models and frontier labs 4) no one else will have such advance access to frontier models or be able to spend as much on tokens as the labs themselves 5) so, it makes sense for frontier labs to scan and patch open-source software, especially the code their coding agents use to build software 6) one issue, though, is that open-source projects often have difficult governance structures, so i can imagine a world where frontier labs just fork these projects 7) build and maintain their own versions, making fixes, improving performance, adding features, etc and then train their models to use them by default 8) i can also see a world, very soon, where ~all software at a company has a claude or codex instance running 24/7 that watches over it (like the anthropic superfans) 9) the agent watches an anthropic or openai feed and, when a vulnerability is found, immediately kicks off a full migration to the new version 10) or otherwise refactors the codebase, changes the infrastructure, ports capabilities, etc... to mitigate the vulnerability 11) this will happen automatically, without a human in the loop; engineers will just come in the next morning and see what happened 12) this is part of what it means for coding agents to increasingly own software, and for labs to become responsible for the software supply chain 13) at some level, i think this just becomes the enterprise expectation when they are paying millions for coding agents / the agents are making arch choices 14) it is a very exciting time to be at a frontier lab and think through what this product offering should look like and how it should be sold to enterprise clients
33
24
1
332
30,334
My gym visits · 5×5
1
1
133
Doing super light weights now at the start, and can really feel the tendons around my knees getting stronger and knee pain going away.
23
Back at it, week 3
55
This article put words on something I see too. LLMs iterating on code written by LLMs tends to degrade over time. Code commens become longer. Metadata becomes self-referential. Tests becomes superfluous. Stale code is maintained instead of deleted.
1
1
80
Alterantively, the cobe base overatkes its original creator, and gets better and better, but the humans can't keep up. Very hard to tell the difference though!
15
More and more often, the questions generate by AI can only be answered by AI. Either this is decoherence, or its a new language being created by Claude.
1
1
86
It's a bit silly that Claude can do almost anything, except using its own tools. It will happily manage sharing permissions on any other tool, but not in a Claude artifact.
2
92
The goal is to create the most comprehensive, beautiful, and fun timeline of the history of the Åland Islands. Ideally in collaboration with other vibe coding amateur historians. We’ll see.
12
So my latest though about coding is that building with LLMs is more fun than playing them. Side projects are more like model railways than infant products. So I’m building something like that, but based on my love of historical timelines and maps. github.com/321k/aland-tidsre…
16