The cookie that stayed behind
Flash-Next builds a cookie store, then misses its simplest exit.

An HTTP client remembers cookies so you don't have to send them yourself with every request. When you clear that store, you expect the next request to leave them behind. In this run, one came along anyway.
Flash-Next's assignment was to give HTTPX a deterministic cookie store: predictable rules for keeping cookies, choosing which ones belong on a request, and sending them in a stable order. The implementation reached into the client as well as adding the store itself. The agent wrote tests and caught problems along the way, including how host-only cookies behaved on subdomains.
Its own 39 tests passed. The separate official grader found one more case: call clear() without arguments, then make another request. The outgoing header still contained a=1.
The cause was small enough to read in one line. The filter kept a cookie if the supplied domain and path were both None. With neither argument supplied, every cookie qualified to stay. The default way to empty the store did the opposite.
After about 12 minutes of agent work, the result was 114 of 115 feature checks and all 1,281 regression checks passed. That's a lot of working behavior beside a very ordinary bug. I'd want this failure spelled out in a code review rather than rounded away as “almost all tests passed.” The exact result and run settings are below.
Add a deterministic CookieStore with modern Set-Cookie parsing
Add a deterministic CookieStore with modern Set-Cookie parsing, cookie eviction, and request header handling.
Run history
Flash-Next IQ3_SDeepSWE · follow-on · 2026-10-12
- Agent time
- 12m 02s
- Feature checks
- 114 / 115
- Regression checks
- 1,281 / 1,281
Run settings & provenance
- Run ID
- tonight-20261012-followon/httpx-deterministic-cookie-store__9F5go7T
- Model
- qwen3.8-flash-next-iq3_s
- Reasoning effort
- medium
- Requested output allowance
- 196,608
- Temperature
- 1.0
- Top p
- 0.95
- Top k
- 20
- Min p
- 0.0
- Presence penalty
- 0.0
- Repetition penalty
- 1.0
- Agent
- OpenCode/Pier
- Runtime
- Strata 0.1.42
- Agent timeout seconds
- 2,700
Timing, delivery & token accounting
- Timing seconds · Agent execution (seconds)
- 722.03391
- Timing seconds · Environment setup (seconds)
- 3.742121
- Timing seconds · Verifier
- 22.777874
- Usage · Input tokens
- 2,198,191
- Usage · Cached tokens
- 2,176,076
- Usage · Output tokens
- 32,084
- Usage · Peak context tokens
- 53,651
- Usage · Steps
- 58
- Usage · Requests
- 59
- Patch · Bytes
- 40,321
- Patch · SHA256
- 6daed644994b555a569adbd025079a8a8df8210e6c164c084a1ba198793bf83b
- Verification · Status
- official-scored
- Verification · Verifier
- DeepSWE v1.1 official verifier
- Verification · Evidence
- verified campaign state and official verifier result
Published comparisons
Published comparison sources
Published successful attempts using mini-swe-agent on the same task, not matched local runs.
gpt-6-astra · xhigh4/4 passed
Published attempt details
- run attempt
- 1
- outcome
- pass
- passed
- True
- run attempt
- 2
- outcome
- pass
- passed
- True
- run attempt
- 3
- outcome
- pass
- passed
- True
- run attempt
- 4
- outcome
- pass
- passed
- True
claude-opus-5 · max4/4 passed
Published attempt details
- run attempt
- None
- outcome
- pass
- passed
- True
- run attempt
- None
- outcome
- pass
- passed
- True
- run attempt
- None
- outcome
- pass
- passed
- True
- run attempt
- None
- outcome
- pass
- passed
- True
Official assignment
Recorded assignment context
Add a deterministic CookieStore with modern Set-Cookie parsing, cookie eviction, and request header handling.