Three stories from the last week that say something about where tooling and platform money are going. One is a popularity index that people read too literally. One is a TypeScript framework making a serious performance claim. One is a Greek bank deciding it wants to own the software layer of agriculture.
Java passes C++ again, and the top 10 shuffles
The November 2026 TIOBE Index puts Java ahead of C++ and drops Rust out of the top 10 entirely. Go takes the vacated slot. Dart entered the top 20 for the first time at 0.74%, up from 27th place in October. Julia, which came back last month, is already gone again.
TIOBE's methodology is search engine queries for language names. It does not measure lines of code, job postings, package downloads or anything a compiler would recognize. Paul Jansen, who runs the index, credits Dart's climb to Flutter, Google's cross-platform framework that uses Dart as its only language. That explanation is plausible and also untestable from the index itself, since TIOBE cannot distinguish a developer looking up "Dart" for a Flutter project from a student looking up dart-throwing.
The Java-over-C++ flip is worth roughly nothing on its own. Both languages sit in a band where month-to-month movement of half a percentage point is noise. I have watched this index swap positions between the same two languages a dozen times across a decade, and no production system has ever been affected by it. If you are choosing a language for a greenfield service in late 2026, TIOBE will not tell you anything that your hiring pipeline and your existing dependency tree cannot tell you better.
Rust leaving the top 10 is a slightly more interesting signal, mostly because it is the kind of headline that gets quoted in budget meetings. Rust is not going anywhere. It is in the Linux kernel, in Android's platform code, in Cloudflare's edge, and in a large chunk of new systems work. What TIOBE measures is how often people type the word into a search box, and Rust developers stopped needing to look things up years ago because the compiler tells them everything.
Effect 4.0 and the numbers nobody has reproduced
Effect 4.0 shipped on 30 September 2026 after eight months of beta and six weeks of release candidate. It is a from-scratch rewrite of the fiber runtime, and the team's own benchmarks are aggressive: a minimal program using Effect, Stream and Schema drops from 35.6 kB to 7.1 kB, a 5x reduction. Heap usage for 50,000 fibers falls from 157.5 MB to 21.8 MB, an 86% cut. Concurrent task throughput goes from 0.71 million to 4.57 million tasks per second, about 6.4x.
Those numbers are medians of nine fresh-process runs on identical source. They have not been independently reproduced, and stork.ai noted as much. Treat the throughput figure as a ceiling rather than an expectation: 4.57 million tasks per second on a synthetic workload tells you the scheduler is cheap, not that your HTTP handler will get faster. The bundle size number is the one I trust most, because gzip output is easy to verify and hard to fake.
The structural changes matter more than the benchmarks. Core modules from @effect/platform, @effect/rpc and @effect/cluster now live inside the main effect package. Every @effect/* package shares a single version number. The core has zero runtime dependencies after removing pg, which was replaced by a native PgProtocol client, and dropping MessagePack. A TypeScript library with no transitive dependencies is a practical gift. I have spent entire afternoons reconciling version conflicts in a dependency graph where one package wanted an old pg and another wanted a newer one.
The LTS policy is the part I would push a team to read closely. Effect 4.x gets bug fixes until September 2029, or one year after 5.0 ships, whichever comes later. Security fixes run to September 2029, or two years after 5.0. That is a real commitment written down, which is more than most open-source projects offer.
Two caveats. First, the launch download numbers are misleading: 43.9 million weekly downloads covers the whole ecosystem across beta and RC tags, while stable 4.0 recorded about 150,000 downloads on its first day. Second, the team admitted after launch that several modules intended to ship as unstable went out without a @stability tag, which implicitly promised stability they did not have. Patches 4.0.1 and 4.0.2 on 8 October added the tags and moved HTTP, SQL, RPC and AI spans to OpenTelemetry semantic conventions. Shipping a stability annotation system and then failing to annotate is a rough look, but the fix was fast and every exported API now declares its status.
Effect 4.1 followed on 20 October with effect/crypto modules for randomness, hashing and signing, a Glob module backed by FileSystem.glob on Node, Bun and Deno, a streaming Csv module in effect/encoding, and RFC 9457 Problem Details support in HttpApi. If you are on 3.x, the migration guide is the starting point, and the team suggests coding agents can do most of the mechanical work. I would not hand a fiber runtime rewrite to an agent without a test suite that exercises error channels and interruption semantics. Typed errors are exactly the kind of thing an automated migration gets subtly wrong.
FarmClick and the bank-funded farm platform
Piraeus Bank and Wikifarmer set up a joint venture called FarmClick to push digital transformation into the agri-food sector. The announcement, reported by Thanasis Koukakis at Dnews, stresses the bank's financial strength and Wikifarmer's marketplace reach. Details on the product surface are thin so far.
The pattern is familiar from other European markets: a bank owns the distribution and the credit data, an agritech company owns the producer relationships, and the joint venture becomes the ordering and payments layer between them. Whether it works depends on whether farmers actually want another portal. Farmers in Greece already deal with cooperative paperwork, subsidy applications and buyer invoices. A platform that replaces some of that is useful. One that adds a login is not.
What I would do with this
If you maintain a dependency-heavy Node or TypeScript service, Effect 4.0 is worth a spike on the dependency count alone. Zero runtime dependencies is rare enough that it changes your security patch cadence.
Ignore the TIOBE month-to-month swaps. Watch the yearly shape instead, and watch Dart specifically, since Flutter adoption is a real thing you can verify with app store listings.
And keep an eye on FarmClick. Bank-backed vertical software has a habit of becoming mandatory infrastructure in markets with few alternatives, which is either good or bad depending on whether the API is documented.