Blaze Next - Vision, Ideas, Tasks etc

Hi everyone :waving_hand:

This thread went quiet after the summer, so here is a catch-up: what happened in 2026, where 3.1 stands, what I think matters now (spoiler: AI changed my list), and three polls so you can weigh in in 30 seconds.

What happened this year

WhenWhat
Feb 19First Blaze user group meeting, @jkuester opens this thread
MarchES2015+ everywhere (#488), zero npm dependencies (#489), jQuery becomes optional with a native DOM backend (#493), fewer DOM writes on attribute updates (#478 by @wreiske), template exceptions on console.error (#495)
AprilBlaze 3.0.3 released, docs migrated to VitePress (#494), whole test suite running on both backends (#497). I also posted an SSG/SSR proof of concept for Blaze, split into blaze#507 and meteor#14355, now scheduled for Blaze 3.2 and Meteor 3.7+
Jun 4Second meeting (recording)
Jun 253.1.0-alpha.0 on Atmosphere
SeptemberI ran the alpha on 11 of my projects with and without jQuery (report), which caught a native backend bug (#512, fixed in #513). The two-year-old stale {{#each}} bug (#468) got fixed too (#501)
Sep 10meteor create --pwa lands in Meteor 3.6 (meteor#14473), and the PWA skeleton is a Blaze app. The offline data layer is the next step, requirements are being collected in this thread
Sep 233.1.0-beta.0 published by the core team (announcement)
Sep 30Lazy-loaded templates with Rspack were not lazy at all; @nachocodoner fixed it in meteor#14805 after a report from thomastraum
Oct 10@drone1 is moving a production app off jQuery and found three bugs in the native backend (#517, #519, #520), with fixes attached (#518, #521, #522). Exactly the kind of feedback a beta is for 🙏

13 PRs merged, 8 still open, and Meteor 3.6 now has its own end-to-end test group for Blaze apps plus a Blaze PWA skeleton. For a thread that started with “is Blaze still a thing?”, I’ll take it.

Where 3.1 stands today

  • 3.1.0-beta.0 is the current version to test. Meteor 3.6 (rc.1 right now) still comes with Blaze 3.0.3, and the plan on the release PR is to get a stable 3.1.0 into Meteor 3.7.
  • drone1’s three fixes are reviewed and green on both backends. Jan is bundling them on one branch, then we get a second beta or an RC.
  • The docs need a pass before the final: the migration guide still says jQuery was removed (it is optional), the API page says this.$() always returns a jQuery object (without jQuery it returns an array), and blazejs.org still serves the old Hexo site while the VitePress one sits on the release branch.
  • One more bug I found on the way: an async helper can keep a stale result when a slow promise resolves after a newer one (#514). It has been there since Blaze 2.7, the fix is agreed with Jan, PR coming.

Everything else (visual error indicator, Blaze.setExceptionLogLevel(), inline expressions, server-side rendering, oxlint) is parked on the 3.2 milestone, waiting for 3.1 to ship. Server rendering is the one I care most about in that list: the Blaze half (#507) only needs a rebase once 3.1 is out, and the Meteor half (meteor#14355) is on the 3.7 track.

What I think matters now

Back in February I wrote: invisible modernization first, then the foundation, then new capabilities. The first part is done. But something else happened this year: most of us now write code with an AI assistant sitting next to us (my own apps are mostly AI-written at this point, and I know I’m not alone here). That changed my priorities more than any Blaze PR did, so here is my updated list. Please disagree in the replies :slightly_smiling_face:

Things I would stop spending time on. An IDE extension for Blaze (syntax highlighting, snippets, go-to-helper). It was on Jan’s list in February and in the June meeting notes, and I understand why. But today Claude or Codex writes the template, the helper and the event map in one go, and they never use the extension. Same for scaffolding tools and generators.

Things I would spend time on instead.

  1. Make Blaze something AI gets right the first time. Blaze is tiny and predictable, which is a huge advantage here, but the rules live in our heads: what is allowed in an attribute, how async helpers behave, when Template.instance() is null after an await, how event delegation works without jQuery. Those rules need to be written down where an assistant can load them: an llms.txt on blazejs.org, the meteor-blaze skill in meteor/agent-skills kept in sync with every release (it exists, and it is exactly the right place for those rules), and compile errors that point at the file and the line (drone1’s point above about templates under /imports is a perfect example). The cheaper it is for an AI to produce correct Blaze, the more Blaze gets written.
  2. Performance. Blaze is fast enough for most pages and slow on big lists. On a test app I measured reversing 5,000 rows in {{#each}} at 11 seconds; with a patch on observe-sequence it drops to 3.4 seconds with identical callbacks, and there is more to gain in @index updates. I want benchmarks in the repo so every PR can show its numbers.
  3. Finish what is started. 3.1 final, then the 3.2 PRs that already exist, server rendering first. On the Meteor side, meteor create --pwa only scaffolds Blaze for now, so the offline layer will land on Blaze first. Expressions and single-file templates are nice to have for me, but they come after the two points above.

And keep the simplicity. It is still the reason I pick Blaze for new projects.

Your turn (30 seconds)

Where are you with Blaze 3.1 and jQuery?
  • Still on 3.0.x, have not tried 3.1 yet
  • On the 3.1 beta, kept jQuery
  • On the 3.1 beta, removed jQuery
  • Tried removing jQuery, went back
0 voters
What cost you time with Blaze this year? (pick up to 3)
  • Confusing compile or build errors
  • Async helpers and await in event handlers
  • Slow {{#each}} on big lists
  • Code splitting or lazy loading templates
  • jQuery plugins and their lifecycle
  • State management boilerplate
  • No expressions in templates
  • No scoped CSS or components
  • No server rendering
  • My AI assistant writes wrong Blaze code
  • Something else (tell us below)
0 voters
What should we work on next? (pick up to 3)
  • Performance and benchmarks
  • AI-ready docs, llms.txt and agent skills
  • Better error messages
  • Inline expressions
  • Single-file templates
  • Server rendering
  • Offline and PWA support
  • Finishing the jQuery-free backend
  • An IDE extension
  • TypeScript types
0 voters

If you have five more minutes: reply with the one thing that broke or annoyed you in Blaze this year. A link to an issue, a snippet or a sentence is perfect, no need for a full report. And if you can spare an hour for a code review, drone1’s PRs are small, tested on both backends, and a great place to start :smiling_face:

@jkuester, up for a Blaze meeting in November to go through the answers? I’ll summarize the polls and replies here before that :+1:

2 Likes