Why Learning Brain ELT exists.

What it actually is

Learning Brain ELT is a plug-in for AI tools. A lightweight service your AI talks to when you ask it for ELT design or audit help. It doesn't replace your AI. It sits alongside, supplying the SLA and materials-design expertise that general-purpose models don't have.

When you ask Claude (or Codex, or ChatGPT) to design an ELT lesson, write a vocabulary set, generate a reading or listening task, or audit a draft coursebook page, your AI connects to Learning Brain ELT automatically. What comes back is three things:

• Curated SLA evidence. The specific research relevant to the task, with cited sources and an evidence-strength rating on every claim.
• A quality rubric. The review criteria a senior ELT editor would apply to this kind of work.
• Instructions for the AI to follow. The structure it should use when producing the output, plus refusal logic for content-shaped goals, mis-pitched material, and pragmatically infelicitous design.

The AI produces the finished work from that package. Learning Brain ELT makes sure the work is structurally sound, level-coherent, and defensible against the rubric a senior commissioning editor would apply.

Seventeen tools cover the main jobs of ELT writing and editing: eliciting the ELT context (CEFR, L1, skill focus, instructional setting), designing single-lesson skeletons, writing vocabulary sets and reading / listening / grammar / speaking tasks, designing synchronous speaking sessions, and auditing draft material. Six expert personas organise them: Cross-cutting, Learning Scientist, Curriculum Architect (ELT), Instructional Writer (ELT), Course Doctor (ELT), and Delivery Coach (ELT).

Behind the tools sits a curated knowledge base of 314 research notes — 195 cross-cutting learning-science notes shared with the core product, plus 119 ELT-specific notes across 15 domains covering SLA foundations (Krashen, Long, Swain, Schmidt, DeKeyser), vocabulary acquisition, grammar instruction, the four skills, language testing, materials and publisher workflows, the CEFR Companion Volume 2020 first-class additions (mediation, plurilingual competence, online interaction, phonological competence), inclusion and accessibility, teacher development, discourse and pragmatics, and specific contexts (CLIL, EMI, EAL, ESOL, business, EAP, ESP, young learners, exam prep, vocational).

You keep using the AI the way you already use it. The conversation doesn't change. The quality of what comes back does.

How it pushes back

Most AI tools for ELT produce surface-plausible material. Dialogue that scans. Reading texts that pitch roughly to a stated level. Vocabulary sets that look reasonable. Comprehension exercises with four plausible options each. Learning Brain ELT is designed to resist that in three specific ways.

1. Evidence only, never invention

Tools cite from the substrate or they refuse. Each substrate note is a complete, cited argument with an evidence-strength rating (strong, moderate, weak, or theoretical). When a tool has nothing substantive to say on a topic, it says so rather than falling back to general SLA knowledge. Popular ELT myths — learning styles in language learning, "younger always learn faster," "native speakers are inherently better teachers" — get flagged the moment they appear.

2. Specific structural refusals

Some requests don't get built, regardless of how confidently they're asked. Content-shaped lesson goals ("cover present perfect") and aspirational ones ("become more fluent") get refused with an explanation. Vocabulary sets that mix wildly across CEFR levels get refused. Reading tasks asking 5 gist items on a single text get refused (gist is a whole-text construct). Natural-fast speech-rate listening tasks at A2 get refused (decoding capacity at the level doesn't support it). Speaking-task prompts dressed as tasks ("discuss travel") get refused. Audits of material under 60 characters get refused. Each refusal is documented; each catches a failure mode that surfaced in testing.

3. Senior-editor-grade rubrics

Ten of the seventeen tools carry a rubric the AI is instructed to answer to before returning anything. Aspirational lesson goals get rewritten, not softened. Mis-pitched vocabulary loads get caught. Reading items whose construct doesn't match their format ("MCQ for gist" where the answer is a lifted clause) get flagged. Distractor rationales that say nothing ("this option might confuse them") get rejected. Every rubric includes explicit anti-agreeableness instructions for the AI: name structural flaws directly, refuse hedge words and compliment sandwiches, keep pushing back when the material is bad.

The first two mechanisms hold regardless of which AI you're using. Rubric compliance is highest on Claude and Codex, which is why those are the recommended clients. ChatGPT chat follows the rubrics less consistently on complex audits. The tools still work there, but the best experience lives in Claude and Codex.

Why a curated substrate

Most AI products that claim to know a domain use a simple retrieval step: take a question, search through a pile of stored documents, hand the model the most relevant chunks. That doesn't work well for ELT, where the primary material is dense applied-linguistics papers, contested methodologies (PPP vs TBLT, explicit vs implicit grammar instruction), and frameworks that evolve over time (the CEFR Companion Volume 2020 added entire scales the 2001 framework didn't have).

Learning Brain ELT uses a different approach. The 119 ELT-specific substrate notes are hand-curated, each written as a complete argument: a defensible claim, the primary research behind it, the conditions under which it holds, and the design implications. Each note cites primary sources. Each carries an evidence-strength rating. Each links to related notes so the tools can reason across topics. When a vocabulary-set question touches lexical-coverage thresholds, the productive-vs-receptive-vocabulary-gap and spaced-retrieval notes are one link away.

The substrate is publisher-neutral. It draws on Oxford 3000/5000, Cambridge English Profile, NGSL, AWL, ALTE standards, CEFR descriptors, and the wider applied-linguistics literature — without privileging any single publisher's output. Where multiple frameworks compete (e.g. word-list approaches, CEFR alignment methodologies), the substrate names the alternatives rather than picking one.

Its relationship to Learning Brain (core)

Learning Brain ELT is a sibling product to Learning Brain (core). The core product covers general learning science and corporate L&D — 32 tools across five expert personas (Learning Scientist, Curriculum Architect, Instructional Writer, Delivery Coach, Course Doctor), built for instructional designers, training managers, and course creators.

Learning Brain ELT shares the cross-cutting and learning-scientist tools (the same 9 lb_* and ls_* tools) but layers ELT-specific design, writing, audit, and coach tools on top, plus the SLA / ELT substrate alongside the shared learning-science substrate. The two products run as separate Fly apps on the same codebase — same Docker image, different pack selector. They can coexist in the same Claude or ChatGPT account; the skill files use distinct names (learning-brain and learning-brain-elt) so they don't collide.

If your work is corporate L&D, instructional design more broadly, or general training-content development that isn't ELT-specific, the core product is the right install. If your work is freelance ELT writing, commissioning editor work at an ELT publisher, teacher-training (CELTA / DELTA / TESOL Cert), in-school EAL, or adult ESOL, Learning Brain ELT is the right install.

How it was built and tested

Learning Brain ELT was built on top of the existing Learning Brain (core) architecture during 2026-05. The 17-tool surface and 119-note ELT substrate were developed against four explicit audiences (freelance writers, commissioning editors, teacher-trainers, in-house publisher editors) with the rubrics calibrated to what a senior commissioning editor would catch on a draft coursebook page.

Testing approach: structural snapshot tests gate every code change (17 tools / 24 domains / 314 notes loaded successfully); rubric self-checks run on every design / write output before returning; the substrate was externally reviewed in 2026-05-26 and substantially expanded in response (62 → 119 notes / 9 → 15 domains across six new domains).

Known gaps: there's no longitudinal study showing that materials designed or audited with Learning Brain ELT produce measurably better learning outcomes than materials designed without it. The structural differences are clear in side-by-side comparisons against a published coursebook page. The direct causal chain from "used this tool" to "learners retained more / acquired faster" takes years and access to publisher production datasets to measure. The substrate is comprehensive across the principal SLA and ELT-publishing literature; it's thinner on emerging frontiers (large-scale AI in language assessment, multimodal pedagogy beyond text-and-image, very-young-learner content beyond Pre-A1 framework guidance). When the substrate is silent on something, the tools say so rather than bluffing.