Indragie Karunaratne retweeted
After building a completely working iOS app with Codex last night it’s pretty clear that there’s absolutely no reason to use React Native anymore in 2026
My year of Linux on the desktop was off to a bad start because the Omarchy ISO wouldn’t boot on my machine with an NVIDIA GPU, but then I realized that I’m living in the future and had Claude patch the ISO (with safe mode graphics) for me. It works!
Opus 5.5 feels like a big step up in codegen capabilities, but I'm finding that it generates the strangest sounding prose. Docs, comments, UI copy, PR titles/descriptions are unusable.
Indragie Karunaratne retweeted
i'll be speaking at @Cloudflare connect next week!
i'll be talking about detecting and preventing prompt injection through @sentry in a cost efficient manner.
a non-security person talking about security! again!
My wife loves watching what our cat is up to at home, so I had Opus 5.5 use captured HTTP traffic from the mobile apps for Ring (cameras), Petlibro (feeder), and Whisker (litter robot) to build an hour-by-hour summary of everything he's been doing.
Ring already uses AI to tag the clips with whether an animal was detected, so I already have a smaller candidate set of clips to look at. I generate still frames every 1s from each clip and pass the sequence to the model.
Indragie Karunaratne retweeted
I wrote a bit about what Codemode actually is and how it works: lucumr.pocoo.org/2026/10/6/c…
Indragie Karunaratne retweeted
Hi friends, lets play a game where we spot the prompt injection attempt!
I promise I'm not using y'all to label data 👀
Indragie Karunaratne retweeted
It’s been a week since I moved one of eval datasets from Gemini 3.1 to Jev by @typesafeai for LLM judging. The results so far:
- No meaningful change in scoring accuracy
- ~200× cheaper (~$0.01 → ~$0.00005 per judge)
- ~50× faster (~10s → ~0.2s median)
- ~50% fewer input tokens, 87% fewer output tokens
But not everything’s perfect, more details in the thread
We’re also hiring engineers to help build our eval platform and enable every product team at Sentry to ship AI features with confidence: sentry.io/careers/95d2eeab-2…
We’re investing heavily in evals at @sentry. It’s early but we’re seeing a culture change from shipping vibes to shipping with clear data using our evals platform.
Will share more in a future blog post, but some stats from our devs in the past month:
- ~1800 eval runs triggered directly on PRs
- 1500 eval scenarios ran across 17 AI surfaces in our product
- Cost/token reductions of ~20-50% across a few of our surfaces
- Eval passing improvements of ~10-20% from some core “code mode” changes launching soon (shoutout @grichadev)
All this from essentially 0 back when we started this effort in May.
Indragie Karunaratne retweeted
i think jev is resonating with devs so well b/c it unlocks so many opportunities for composing ai into systems and products rather than ai _becoming_ the product/system
really does feel like it was a missing primitive
I used the first Dropbox beta back in 2006 and was sold right away - I’m still a paying customer of the product today, 20 years later. But there hasn’t been much innovation in this space since then and we’re in the midst of a broad shift in how people interact with computers.
I’m excited to back this great team and see what a modern take on this problem looks like!
Space has raised a $2.4M pre-seed to give your computer infinite storage.
Today teams lose hours each week downloading files before humans and agents can begin working.
Space removes the wait. Every file, instantly accessible while using zero disk space.
Indragie Karunaratne retweeted
it will be much faster & MUCH more capable, coming sooooon
Code mode's having a moment (some more litigiously than others). We're building it into Seer @sentry. It's faster, cheaper & acts for you.
Pointed it at 54 JS issues → triaged them all, clustered dupes, archived the 20 noisy ones with a comment each. Shipping soon 😇
Indragie Karunaratne retweeted
we have metrics btw
and logs
and session replay
and continuous profiling
oh yeah and we can help you debug your agents too
Sometimes when you do one thing extremely well (error monitoring), it's hard to change perception that you do more things now (@sentry has 10+ products). can BLG (billboard-led growth) fix this? time will tell. ⏱️🤔




























