Jakob Norlin

We just opened up Endform to public access 🎉 Come join the fun and run your Playwright tests faster!

seems obvious but: things that are changing rapidly: 1. context windows 2. intelligence / ability to reason within context 3. performance on any given benchmark 4. cost per token things that are not changing much: 1. humans 2. human behavior, preferences, affinities 3. tools, integrations, infrastructure 4. single core cpu performance therefore, ngmi: 1. "i found this method to cut 15% context" 2. "our method improves retrieval performance 10% by using hybrid search" 3. "our finetuned model is cheaper than opus at this benchmark" 4. "our harness does this better because we invented this multi agent system" 5. "we're building a memory system" 6. "context graphs" 7. "we trained an in house specialized rl model to improve task performance in X benchmark at Y% cost reduction" wagmi: 1. product/ui 3. customer acquisition 4. integrations 5. fast linting, ci, skills, feedback for agents 6. background agent infra to parallelize more work 7. speed up your agent verification loops 8. training your users, connecting to their systems and working with their data, meeting them where they are

@thdxr @RhysSullivan Fake channel

we made his dreams come true

@NathanFlurry Ah okay too bad! Was really interested in it, but for our cases it has to support all npm packages. Still, really cool tech!

@cakemakerjake currently not possible. banking on atwood's law.

Rivet Secure Exec 0.2.0 now has 2.8x faster coldstarts for Node.js execution without a sandbox 17.9 ms → 6.2ms (p99) 16.2ms → 4.0 ms (p50) Instead of rebuilding the runtime every creation, V8 isolate snapshots let us clone a ready-to-go instance instantly – much like in KVM

Fun things coming up…

Long but fun day shooting a video. Announcement coming soon!

@cnakazawa @youyuxi @birch_js @boshen_c 👀

@cakemakerjake @youyuxi @birch_js @boshen_c Good news, you don't need `setup-node` with Vite+, just use github.com/voidzero-dev/s…

Vite+: We'll be adding support for Bun as a package manager, but we are not planning on making bun work as a runtime at this time. github.com/voidzero-dev/v…

@youyuxi @birch_js @cnakazawa @boshen_c Only reason I'd want it to run in Bun is because the setup-node GH action takes like 20s vs setup-bun's 3s.

@birch_js @cnakazawa @boshen_c Also - due to (1) our tools have wider use case coverage than Bun’s and (2) most of our tools run in native Rust, the benefit of switching from Node to Bun during dev is negligible, or arguably negative (due to potential behavior divergence)

@boristane Signal > RCS > WhatsApp > SMS > iMessage

whatsapp > imessage

@MichaelArnaldi I guess I recorded macros, not wrote them

@MichaelArnaldi Loved writing macros for repetitive tasks!

That's for kids, if you recognise this you're old

@boristane @thdxr serverless server less less server lé server server

@thdxr serverless

i thought i was out i stayed out of the game for 7 years but here we go again

sent this to the team today everything great comes from being able to delay gratification for as long as possible and it feels like we're collectively losing our ability to do that

@kylegalbraith Big news, congrats!

@_ashleypeacock I'm genuinely curious, because I like the thought of designing a service like this, but fail to see how it would practically work.

@_ashleypeacock This makes sense when you want to show a page with info for a specific stock, but how would this work if you for example want to show a list of all stocks, and make it sortable for different properties? You'd have to make a request to every single DO and sort in memory, right?

Durable Objects make your applications feel alive in a way I've never experienced before, it feels very novel. To me, it feels like the technological equivalent of watching a group of dancers on stage. Let's look at a concrete example: say you want to monitor every stock in the world, from the US to the UK to Australia. Some data might be pulled (via APIs), some data might be pushed (via webhook). Users might want to view live prices, historical aggregate data, or sign up for price alerts. Without Durable Objects, you need all sorts to make this a reality - load balancers, caches, databases, a way to handle WebSockets, scheduling as well, and some form of parallelism. With Durable Objects, they can do everything and manage the whole lifecycle: - Each Durable Object represents one stock, with its own compute & storage - Using alarms, it can be woken up on a schedule to retrieve the latest data (e.g. every minute) - Alarms can be set intelligently, only triggering when the market is open - Webhooks are even easier, just have a Worker route to the appropriate Durable Object - You get load balancing for free. Workers scale up as needed. - Price alerts can be handled again using alarms, or as separate DOs (see below) - Intermittent API errors are handled gracefully, as alarms retry with exponential backoff. Errors are isolated to one Durable Object. - For live UI updates, clients can connect to Durable Objects using WebSockets - Aggregating data is rapid, thanks to 0ms latency SQLite-backed Durable Objects If you tried to build this without Durable Objects, you'll probably spend most of your time with configuration and infrastructure. Durable Objects are defined as classes in your code, with methods to implement all the features above out of the box.

@thdxr Setting your messaging up for secure containers? 😶‍🌫️

we'll adjust to this new world but this is why i was saying the people responsible for security/safety are freaked out your company maybe lax now to let everyone experiment but those times are probably ending

@NathanFlurry @rivet_dev You mean they changed that at the same time as well?

@cakemakerjake @rivet_dev Context for the post x.com/zebassembly/st…

PSA: Cloudflare & Durable Objects run "near" your AWS datacenter, not in it. This means you're getting. 1. Higher latencies 2. Paying for ingress fees 3. Worse p99s → jitter & dropped packets @rivet_dev (open-source Durable Objects alt) can run on AWS natively

@thdxr @kitlangton why is there try/catch in there?

we've had a process where any PR can be labeled as "beta" and gets applied to beta builds this wasn't practical before because you'd end up with a bunch of merge conflicts but we just ask opencode to resolve conflicts in those cases and we haven't had a single failure since