a hungry orb that eats your agent’s context

it eatsyour context.

try it: scroll to feed the orb · move your cursor and it follows

100% of creator rewards route into moonah liquidity sweeps. Powered by @pumpfun: moonah.solpads.fun

Buy $MOON on pump.fun →DexScreener ↗cost story ↓

$0.20per 1M cache reads. same on both models.
1.88x → 1.26xopus / sonnet per turn, 20K → 400K.
$1.50to switch models at 300K. a 20K handoff: $0.10.

pump.fun ↗dexscreener ↗solscan ↗

nom

$moon

Ga8n...ibtQ

scroll to feed me ↓

every turn, your agent
re-reads all of this.

user the login flow logs people out at random. fix it, keep the tests green.

tool_use read_file src/auth/session.ts tool_result four hundred lines of session logic, refresh timers, a cookie helper and one very suspicious setTimeout.

assistant the refresh races the logout. if the token expires mid request, the retry fires after the cookie is cleared and the user gets bounced.

tool_use grep "refreshToken" tool_result nine matches across middleware, the api client, two hooks and a test that nobody has run since spring.

tool_use run_tests auth tool_result three failing, one flaky, the rest asleep. stack traces, snapshots, a wall of yellow warnings.

assistant adding a single flight lock around refresh so concurrent calls wait for one promise instead of racing each other into the void.

tool_use edit_file src/auth/session.ts tool_result diff applied. forty lines in, twelve out, one comment that says do not touch.

tool_use run_tests auth tool_result green. green. green. one flaky test still flaky, as is tradition.

this agent execution stream is snacks for moonah. scroll to feed it · scroll back up (or hit spit it out on the meter) to restore the words.

01cost story · keep scrolling to grow the cache

three beats. same models, same price sheet. watch the “2x” gap melt.

context size20.0Ktokens

1.88x

how many times more an opus turn costs vs sonnet

step 1 · 20Kfresh session. opus really is ~2x here — most of the turn is new tokens + output, where opus costs double.

step 2 · 150Kdeep in the session. old context is a cache read at $0.20 / 1M on both models. the gap is melting.

step 3 · 400Kmost of every turn is old context, billed the same on both. opus only needs ~21% fewer turns to cost the same.

sonnet

opus

re-reading old context writing new bits model output

example turn: 5,500 new tokens + 1,500 output · 5-min cache · supplied pricing assumptions.

cost calculator

how: drag the sliders. left = your turn shape. right = live bill for sonnet vs opus.

how big the conversation is new stuff each turn model output each turn

turns if sonnet turns if opus

Assumed rates per 1M tokens from the supplied HTML: sonnet 5.5 $2 input / $10 output / $2.50 write (5m) / $4 write (1h) / $0.20 read; opus 5.5 $4 / $20 / $5 / $8 / $0.20. These are not verified live rates. Check provider pricing ↗

Edit pricing assumptions / reset calculator

opus costs this many × sonnet (per turn)

sonnet 5.5

opus 5.5

re-read old write new output

10Kcontext in cache (log)1M

Move over the chart to inspect the ratio.

total · sonnet

total · opus

the escalation tax

caches are per model. switch a running session at and opus re-writes all of it before reading a single file.

switch with the transcript

clean 20K handoff

the coffee-break tax

default cache lives 5 minutes. walk away longer and the next turn re-writes the whole context.

one rewrite · sonnet

one rewrite · opus

snorb a wallet

how: paste any Solana address or use Phantom to analyze public context load. Evaluates state over public RPC; zero signatures required.

public address only. never asks you to sign. never touches keys. Live requests send the address to your selected RPC provider. Browser CORS or provider rate limits may prevent lookup.

RPC settings · optional Helius endpoint

Use a trusted HTTPS Solana JSON-RPC URL. Works with default public node or your custom Helius/QuickNode RPC endpoint.

check your own logs

how: drop a usage.jsonl (one object per API call). we compute $/turn and $/pass per model in your browser — nothing uploads.

drop usage.jsonl or tap to pick a fileone JSON object per line · fields: task, model, passed, usage

Supported fields & cost assumptions

usage.input_tokens (uncached), output_tokens, cache_read_input_tokens, cache_creation_input_tokens. Also accepts prompt_tokens / completion_tokens. Optional cache_creation.ephemeral_5m_input_tokens and ephemeral_1h_input_tokens. Explicit cost_usd (or cost) overrides inferred cost. Default log cache write TTL is 5m unless cache_ttl is 60 or 1h. Model names containing sonnet or opus use the editable calculator rates; unknown models need explicit cost. Passed tasks are counted once per model + task, not once per call. Failed/unlabelled task costs remain in the $/pass numerator.

results land here. read them in this order: $/turn ratio → avg turns ratio → $/pass.

the expensive model is not
always the expensive task.

sonnet turn thirty one. still circling the same three files, reading the same diff again, hoping it changed.

handoff.md the task, the failing check, the three files that matter. nothing else. no transcript, no dead ends, no forty turns of guessing.

opus fresh session, small context, reads the evidence, writes the fix, runs the check, done in a handful of turns.

snorb burp.

cheat sheet

rules reproduced from the supplied page · test against your provider’s actual billing and workload.

price turns, not tokens.cost per task = cost per turn × turns to finish.

output is the real 2x.thinking is billed as output. heavy output drags the ratio back toward 2x.

long sessions shrink the gap.old context is a cache read at the same $0.20/1M on both.

escalate early.pick the model while the context is still small.

escalate clean.new session + short handoff: the task, the failing check, the 3 files that matter.

pause a lot? 1-hour cache.running straight through? don't.

don't flip effort mid-session.changing top-level effort per turn breaks the cache.

judge by $/pass.“opus feels better” is not a routing rule.