Tell HN: One SWE-bench-Live task: $47 Opus failed, $1.46 GPT-5.6 passed

tamnd1 pts1 comments

tomo-labs/docs/content/experiments/2026/07/13/14-55-dynaconf-doors-closed-lessons-for-tomo.md at main · tamnd/tomo-labs · GitHub

//blob/show" data-turbo-transient="true" />

Skip to content

Search or jump to...

Search code, repositories, users, issues, pull requests...

-->

Search

Clear

Search syntax tips

Provide feedback

--><br>We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Cancel

Submit feedback

Saved searches

Use saved searches to filter your results more quickly

-->

Name

Query

To see all available qualifiers, see our documentation.

Cancel

Create saved search

Sign in

//blob/show;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up

Appearance settings

Resetting focus

You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

{{ message }}

tamnd

tomo-labs

Public

Notifications<br>You must be signed in to change notification settings

Fork

Star

FilesExpand file tree

main

/14-55-dynaconf-doors-closed-lessons-for-tomo.md

Copy path

Blame<br>More file actions

Blame<br>More file actions

Latest commit

History<br>History<br>History

81 lines (63 loc) · 6.33 KB

main

/14-55-dynaconf-doors-closed-lessons-for-tomo.md

Copy path

Top

File metadata and controls<br>Preview

Code

Blame

81 lines (63 loc) · 6.33 KB

Raw<br>Copy raw file<br>Download raw file

OutlineEdit and raw actions

title<br>What the closed-door dynaconf runs teach tomo

linkTitle<br>dynaconf closed-door lessons for tomo

description<br>Seven honest runs on one task, one pass and six fails, read together. The lessons that transfer to tomo: a broad edit that regresses a green test is worse than no edit and wants a do-no-harm gate, spend does not track progress, cache-read is where the money actually goes, and the reflexive git-history probe is fine once and a runaway if repeated. This is the cross-trace synthesis, with pointers to each run.

date<br>2026-07-13 14:55:00 +0700

Seven models ran dynaconf__dynaconf-1225 with both answer doors shut: the git-history door pruned and the network denied.<br>One solved it, six failed, and every one of the seven was honest, no answer fetched.<br>Read as a set they say more than any single run, and most of what they say points at concrete tomo work.

The runs

Model<br>Harness<br>Files<br>Tokens<br>Cost<br>Verdict

gpt-5.6-luna<br>codex<br>25<br>10.1M<br>$1.46<br>PASS, clean

gpt-5.6-terra<br>codex<br>23<br>3.2M<br>$1.82<br>PASS (honest, false leak flag)

gpt-5.6-sol<br>codex<br>23<br>3.2M<br>$2.56<br>PASS (honest, false leak flag)

gpt-5.4-mini<br>codex<br>4.83M<br>$0.78<br>FAIL, clean

gpt-5.5<br>codex<br>19<br>6.08M<br>$4.49<br>FAIL, clean

sonnet-5<br>claude<br>22<br>23.5M<br>$10.32<br>FAIL, clean

opus-4.8<br>claude<br>23<br>20.7M<br>$47.18<br>FAIL, regressed a green test

Lesson 1: a broad edit that breaks a green test wants a do-no-harm gate

The naive read of the table is that breadth or spend decides the outcome.<br>Both are wrong.<br>luna went the widest, twenty-five files, and passed.<br>opus spent the most, $47, and produced the worst result of the seven, a repo where a test that was green at the base commit is now red.

What actually separated the pass from the fails is narrower: did the broad refactor carry the identifier all the way through the loader stack, and did it leave the already-green tests green.<br>luna's did both.<br>opus's did neither, and the second failure is the instructive one.<br>It did not merely fail to fix the target, it damaged working behavior on the way to failing.

tomo already has a convergence guard that stops it running away searching for a fix.<br>It has nothing that stops it shipping an edit that regresses a test that was passing.<br>The gap this run exposes is a do-no-harm gate: after an edit batch, run the in-repo tests the change plausibly touched, and treat a green-to-red flip as a stop-and-reconsider signal rather than continuing to pile edits on.<br>It is model-independent, it would have caught opus before its twenty-third file, and it would not have fired once on luna.<br>That is the property to want: it punishes net-negative edits, not breadth.

Lesson 2: spend does not track progress

Cost on this task ranged from $0.78 to $47.18 and told you nothing about the verdict.<br>The three cheapest passing runs, the gpt-5.6 family, cost $1.46, $1.82, and $2.56.<br>The most expensive run failed and regressed.<br>Turn count says the same: opus took 194 turns and sonnet 235 to reach a wall the gpt-5.6 models cleared in a few dozen tool calls.

For tomo the takeaway is not "be cheap for its own sake," it is "do not read your own token spend or turn count as evidence of progress."<br>A loop that has spent a lot has not thereby earned anything, and the guards should measure movement toward a passing fix, not effort.<br>tomo's leanness pitch holds on the merits here: its honest fails are cheap, and cheap-and-honest beats the $47 fail every time.

Lesson 3: cache-read is where the money...

tomo read file green opus honest

Related Articles