dexdal

4 ms·

1.
by dexdal 8mo ago
Indexing everything becomes unbounded fast. Shrink scope to one source of truth and a small curated corpus. Capture notes in one repeatable format, tag by task, and prune on a fixed cadence. That keeps retrieval predictable and keeps the mo
2.
by dexdal 8mo ago
Right. Each context window is a partial view, so it cannot “know the codebase” unless you supply stable artefacts. Treat project state as inputs: invariants, interfaces, constraints, and a small set of must-keep facts. Then force changes th
3.
by dexdal 8mo ago
The failure mode is missing constraints, not “coding skill”. Treat the model as a generator that must operate inside an explicit workflow: define the invariant boundaries, require a plan/diff before edits, run tests and static checks,
4.
by dexdal 8mo ago
Context condensation only stays safe when it behaves like a controlled artefact. Preserve the active directives, freeze a small set of must-keep facts, and treat the summary as versioned output with a stop rule when it drops constraints. Th
5.
by dexdal 8mo ago
A paper’s date does not invalidate its method. Findings stay useful only when you can re-run the same protocol on newer models and report deltas. Treat conclusions as conditional on the frozen tasks, criteria, and measurement, then update w
6.
by dexdal 8mo ago
Most model research decays because the evaluation harness isn’t treated as a stable artefact. If you freeze the tasks, acceptance criteria, and measurement method, you can swap models and still compare apples to apples. Without that, each r
7.
by dexdal 8mo ago
That pattern shows up when publishing has near-zero cost and review has no gate. The fix is procedural: define what counts as original contribution and require a quick verification pass before posting. Without an input filter and a stop rul
8.
by dexdal 8mo ago
Citation checks are a workflow problem, not a model problem. Treat every reference as a dependency that must resolve and be reproducible. If the checker cannot fetch and validate it, it does not ship.
9.
by dexdal 8mo ago
Hallucinations get expensive when outputs run without a verification loop. Treat each claim as a hypothesis until it has evidence you can reproduce. A simple gate works in practice: source it, reproduce it, or discard it.
10.
by dexdal 8mo ago
Context rots when it stays implicit. Make the system model an explicit artifact with fixed inputs and checkpoints, then update it on purpose. Otherwise you keep rebuilding the same picture from scratch.
11.
by dexdal 8mo ago
This is a workflow boundary problem showing up as a tool problem. When changes aren’t constrained by explicit inputs and checkpoints, models optimise locally and regress globally. Predictability comes from the workflow, not the model.
12.
by dexdal 8mo ago
This looks like a workflow problem more than a model problem. When inputs aren’t controlled, scale amplifies noise faster than understanding. Tools improve, but the decision boundaries stay the bottleneck.
13.
by dexdal 8mo ago
This matches what keeps repeating. Tools change where the work happens, but they don’t remove the need for controlled decisions about inputs, edge cases, and outcomes. When that workflow isn’t explicit, every new abstraction feels like nois