In September 2026, the Debian Code Search maintainer reports removing the last cgo dependency from the project by rewriting TurboPFor integer compression/decompression in native Go on top of Go 1.26's experimental simd/archsimd package. The post says the Go implementation with AVX512 kernels and positional popcount beats DCS's previous cgo-integrated C path, while an equally optimized C TurboPFor baseline remains about 1.4x faster than Go.
Constraint shift
The binding constraint changes from 'SIMD in Go requires cgo, hand assembly, or assembly generation' to 'Go can express SIMD through an experimental standard package, provided the build uses GOEXPERIMENT=simd, targets amd64, and accepts an unstable API.' This shifts the cost of adding a native high-performance codec: escape-hatch dependencies and CPU-specific assembly can be replaced by pure Go, but adoption now carries an experimental-toolchain risk.
Evidence
The document is a maintainer's self-report rather than an independent evaluation. It supplies commit hashes and a benchmark harness but no raw benchmark tables or logs; it names the GOAMD64 level, benchmark flags, and machine pinning, which makes the claims partially reproducible. The article itself qualifies the headline speed advantage: after backporting the same AVX512 kernels and positional popcount technique to C, Go is reported about 1.4x slower. Absent raw data or independent reruns, the top-level 'beats C' framing should be treated as evidence-limited.
A Stigler Center paper, reported through Cory Doctorow's commentary, uses 9,500 acquisition filings with over $1 trillion in customer-relationship intangibles plus earnings-call admissions to show that firms deliberately build lock-in. Sectors with stronger lock-in are reported to have higher margins, fewer entrants, less innovation, and weaker job and wage growth. The work challenges Chicago-school assumptions, but the underlying study is not yet independently assessed.
Constraint shift
The article shifts the central competition-policy constraint from 'dominant firms cannot sustain abuse because dissatisfied customers will defect' to 'firms can make defection structurally impossible, and that lock-in itself deters entry'. If the reported findings hold, enforcers should target switching costs and exit barriers as first-order competitive harms rather than waiting for observable price effects or market entry.
Evidence
The source is a persuasive secondary account, not the underlying paper: it describes a large quantitative dataset and several executive quotations but gives no methodology, controls, or robustness tests. The few illustrative cases are selected to fit the argument, and the headline relationships are phrased in correlational terms in the blog. Treating the study's conclusions as established requires reading the original paper and independent replication.