Alvar Lagerlöf @alvar.dev · 02/10/2026This section header design screams LLM and now I see it everywhere. Sometimes the subheader really really doesn't add value. 210
Alvar Lagerlöf @alvar.dev · 30/07/2026I'm in Jakarta, Indonesia. It's hot, but I've never seen temperature this consistent across days before. It's not actually rainy btw. 170
Alvar Lagerlöf @alvar.dev · 17/04/2026"Electric bus" is "Elbuss" in Swedish, but my silly brain keeps reading it as "El bus" 220
Alvar Lagerlöf @alvar.dev · 27/02/2026No skiing vacation is complete without watching Alaskan men being overly dramatic about driving in winter conditions. 030
Alvar Lagerlöf @alvar.dev · 22/02/2026TIL that if you skip GitHub actions workflows using something like github.com/dorny/paths-... instead of paths:, they are still marked as successful, and can be required checks. This is huge for monorepos. 130
Alvar Lagerlöf @alvar.dev · 15/02/2026At some point, the bus station displays here in Stockholm must have switched to e-link. Or at least that's what I think it is. 000
Alvar Lagerlöf @alvar.dev · 11/10/2025Right down to the more yellowish bottle in the background being slightly taller. 100
Alvar Lagerlöf @alvar.dev · 28/09/2025Hey @danabra.mov, why does this work in Lean? Asking for a friend (actually) 150
Alvar Lagerlöf @alvar.dev · 12/08/2025Looking back, it seems like iOS 11 was the first version that almost completely had good contrast in Control Center. iOS 10 on the left. 110
Alvar Lagerlöf @alvar.dev · 01/08/2025Might be the most subtle rate limit warning I ever got 😅 Didn't ask anything time-related. 010
Alvar Lagerlöf @alvar.dev · 20/07/2025This was kind of a thing in Sweden. It was called Blåvitt (blue white). 020
Alvar Lagerlöf @alvar.dev · 09/07/2025First time trying steamed rice rolls. So good that I left a review immediately. 230
Alvar Lagerlöf @alvar.dev · 08/07/2025You know when you see a random restaurant across the street and it turns out to be a bit? 230
![loss = model(input_ids=inputs["input_ids"], labels=inputs ["input_ids"]).loss
ppl = torch.exp(loss)
print(f"Perplexity: {ppl.item():.2f}") >> Perplexity: 4.26](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:xoabsyepzhmn7jiksspa53ai/bafkreidzgskie2khfnx5l37r4yu3lgjbq2wutjpcjctyxbhardh3s3t4bq)
