SEMAPRAX Tool Links Semantic Code Patches to Verifiable WebAssembly Artifacts

·1 views

Wavect GmbH is developing SEMAPRAX, an experimental open-source agent-native systems programming language currently at v0.2 pre-alpha stage, designed to bring verifiability to code changes made by AI coding agents. A read-only command called 'semaprax target-evidence' rebuilds both original and patched versions of a program and generates deterministic compiler outputs, including a semantic graph, capability manifest, C11 source, and a validated WebAssembly module. Each output artifact receives a domain-separated cryptographic digest, allowing independent verifiers to confirm exactly which projections a given source edit actually affects. The tool uses wasmparser 0.256.0 to structurally validate emitted WebAssembly modules, though the developers stress this confirms only structural validity and not runtime correctness, ABI compatibility, or cross-runtime conformance. SEMAPRAX's current WebAssembly support is intentionally limited to bounded scalar exports and JS/TS bindings, and the project explicitly cautions against treating static evidence reports as stronger runtime guarantees.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Why AI Models Cannot Simply Delete Your Personal Data on Request

When personal data is used to train an AI model, it does not remain as a discrete, removable record but instead becomes distributed across billions of numerical parameters throughout the model's weights. Privacy laws like the GDPR's right to erasure were designed around traditional databases where data can be located and deleted, an assumption that does not hold for trained neural networks. The only guaranteed method — retraining the model from scratch without the requested data — is prohibitively expensive, costing potentially millions of dollars per run and taking weeks to complete. Researchers are developing 'machine unlearning' techniques, such as gradient ascent and influence functions, that attempt to make a model behave as if it never saw specific data without full retraining. However, these methods are approximations and cannot offer the same provable guarantees as complete retraining, leaving a significant gap between legal data deletion rights and current technical reality.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Startups Should Structure Content Moderation Into 7 Risk Categories

A technical framework recommends that startup apps organize content moderation into seven risk categories: harassment, sexual content, self-harm, violence, illegal activity, spam, and PII. Each category alone should not determine an outcome; instead, decisions must factor in severity, confidence, and the intended system action. The design proposes a two-lane pipeline, where high-risk checks like self-harm and credible violence run synchronously to block harmful actions before they execute, while lower-risk quality issues are handled asynchronously. A key principle is that category labels describe content, but policy governs what action follows, since identical flagged content can warrant different responses depending on context and destination. The framework also cautions against over-engineering, advising teams to maintain a single policy object and decision function rather than building multiple unrelated filters.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How a silent PrestaShop hook failure hid three stacked bugs across releases

A PrestaShop module hook registering correctly and showing in Design > Positions still produced no output on the front end, with no errors, logs, or stack traces to indicate why. The root cause was PrestaShop's Hook::callHookOn() silently swallowing all exceptions when debug mode is off, returning an empty string for any failure. Adding a manual try/catch block with explicit logging finally revealed that the Smarty template was failing to compile entirely. The culprit was a plain CSS rule containing curly braces, which Smarty misread as a template tag and choked on, causing the whole template — including unrelated blocks — to silently fail. Wrapping such CSS in Smarty's built-in {literal} tags resolves the conflict and prevents the silent compilation failure.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Dictionary Pattern Matching Silently Ignores Extra Keys, Risking Hidden Bugs

In languages like Python and Rust, dictionary pattern matching does not require an exact structural match — extra keys in a dictionary are silently ignored when a pattern is applied. This design choice prioritizes flexibility, allowing dictionaries with additional keys to match a given pattern without raising errors. However, developers familiar with stricter sequence pattern matching may incorrectly assume the same enforcement applies to dictionaries. In security-sensitive contexts such as financial systems, this silent behavior can allow unexpected or malicious data to pass undetected. Experts recommend that developers carefully review language documentation and implement explicit key validation where strict dictionary shapes are required.

0 comments Read more at DEV Community