#deep-dive
Every post tagged "deep-dive".
107 posts · showing 61–70
-
Why Benchmarks Keep Rising While Deployments Stay Hard
Both reports are accurate. Benchmarks test well-posed, self-contained, checkable problems — and almost none of deployment difficulty is any of those things.
-
The Reliability Ceiling and What Would Have to Change
98% per step sounds excellent and produces 67% on a twenty-step task. The plateau is compounding arithmetic — which means the fix is shortening the chain, not improving the steps.
-
Longer Context Windows Won't Fix Long-Horizon Tasks
The premise is right — things fall out of context — and the conclusion doesn't follow. Most long-run failures aren't about storage, and a bigger window removes the discipline that was helping.
-
What Gets Cheap Next, and What Stays Expensive
Costs are being rearranged, not reduced. Four things are getting cheap, six aren't — and a fourth category nobody plans for is getting more expensive, including maintaining everything that was cheap to create.
-
Predicting Agent Capability: Reason From Mechanism, Not Vibes
"Will agents get better?" is the wrong question. Six things block any given task, only two of them move with model releases — and the other four are the ones nobody's working on.
-
Novel Problems vs Recombined Ones: The Line That Matters
On genuinely novel problems you don't get an error — you get the nearest known pattern applied confidently to an adjacent problem. One classification decides where your attention belongs.
-
Negotiation, Politics, and the Work That Isn't in the Ticket
It's usually called politics, said with distaste, and treated as the tax on real work. It's also the growing share of the job — and most of it is information flow and reliability rather than manipulation.
-
Ambiguity Resolution Stays a Human Job for a While
An agent doesn't fail on ambiguity — it absorbs it, picks a reading, and produces a confident answer to a question nobody asked. Better models guess better; they don't guess less.
-
The Meeting an Agent Can't Attend
Requirements aren't transmitted in those rooms, they're constructed — out of inputs that exist nowhere beforehand and mostly aren't recorded after. And the efficiency instinct cuts exactly the wrong thing.
-
Why Agents Struggle Where There's No Feedback Signal
Coding agents work well because compilers exist, not because code is easy. Effectiveness tracks whether a domain has a fast, independent, specific verdict — and that's frequently something you can build.