neatHack 2026 · For judgesCheck Every FirstRun Claim in 60 Seconds

Tick the 7 evidence rows
Watch the 2:59 demo
See v1 against v2
Open a real receipt Receipts re-issued Oct 11 from saved run state.

EvidenceBack up each claim in your submission with evidence that judges can check.

All 7 Official Evidence Rows Have Proof You Can Open

1The agent improved
In repo
A run before your change and a run after it, e.g., two neatlogs traces, with the numbers you compared
2 of 4 finished, then 3 of 4. 8 traces
2You used neatlogs
Screenshots
Links to neatlogs traces from your project, or screenshots of them
10 dashboard screenshots. See them
3You used Entire
Public
Checkpoints or entire-graph output from your repository
91 of 132 commits carry a checkpoint. the checkpoint commit
4You used cfo.ai
Public
A business plan covering your agent's pricing, costs, revenue and runway
All four, in a public model. cfo.ai model
5The agent recovers from failures
In repo
A run where an action failed, showing what the agent did next
Step 1 failed, then 4 of 4 passed clean. the recovered run 5520c5cf
6You built in public
Public
Links to what you shared during the hackathon, tagged #neatHack
11 tagged posts, 9 after kickoff. See them
7You wrote the code during the hackathon
Public
A public repository with commits dated during the hackathon
First commit 12:34 IST, kickoff 12:30. 96541a8
Trace pages need a neatlogs login. Read the recovered trace in the repo, and screenshots stand in for the rest.

30% The agentHow well the agent, or the system of agents, handles a substantial task from start to finish.

The Agent Runs a Quickstart Start to Finish, and v2 Beat v1

Same 4 quickstarts, v1 against v2

Finished, of 4

v12

v23

Failed steps recovered

v10

v23

Fixes verified clean

v10

v21

Cost per run

v1$0.0262

v2$0.0200

-24%

8 neatlogs traces: before_after_traces.json

Recovery in trace 5520c5cf

step.1failed: curl: command not found
recover.1installed curl
step.2failed: ~/.local/bin not on PATH
recover.2added it to PATH
verifyfresh container, 4 of 4, no model calls

Exported trace

Show details

Every part of a run points to code

Plan: copies each docs command. planner.py
Tools: Firecrawl, Docker and Jev. classify.py
State: saved after every step. state.py
Context: reads only the quickstart page. planner.py
Run: one fresh container per run. executor.py
Recover: classify, patch, retry, 3 tries. recovery.py
Verify: second container, no model calls. verify.py
Ask once: a human is asked only below 0.7 confidence. recovery.py precedents.py

All six numbers

Metric	v1	v2
Finished, passed or verified	2 of 4	3 of 4
Failed steps recovered	0	3
Fixes verified clean	0	1
Tool calls per run	4.25	6.0
Cost per run	$0.0262	$0.0200
Run time, verify included	128 s	201 s

List price, not billed. The fix is commit 3ca6f11. Trace IDs: neatlogs.md. The trace status reads ERROR because a failed step rolls up, but the run ended verified.

25% ToolsHow effectively you used neatlogs, Entire and cfo.ai to build, improve and plan a business around the agent.

neatlogs Improved the Agent, Entire Recorded the Build, cfo.ai Planned the Business

neatlogs

A 52.8x spike led to fix 2de852f.

neatlogs.mdTraces need a login

N1
One run is one trace
N2
A recover.3 span retries a step
N3
Investigate: 52.8x spike, cause at 66%
N4
Detection "docs step broke": 78 flagged spans
N5
Flagged spans, 40 at capture time
N6
Detection analysis over 132 traces
N7
Analytics: 132 traces, $0.9071 total cost
N8
AI judge on trace 5520c5cf: right step
N9
AI judge on audit 4e936e06: 5 of 5
N10
Bisect trace: jafgen 0.4.6 first bad

Entire

91 of 132 commits carry a checkpoint.

Checkpoint, no login graph.ndjsonTrails need a login

9137

91 with a checkpoint
4 setup commits
37 site commits
git log on main up to f5d73eb, 2026-10-11

E1
Commit 3a0dcad carries its checkpoint
E2
Trail 3 blocked merge on a high finding
E3
Trail 3 review: 1 high, 1 medium
E4
Inline review on recovery.py line 228
E5
Entire Trails board with trail #1
E6
Entire analytics after the first checkpoint

cfo.ai

The old $0.50 cap runs out of cash in month 14.

cfo.ai model, no login BUSINESS_PLAN.md

Free

Cap a run: $0.01355
Margin: no revenue

Team

$49

Cap a run: $0.04697
Margin: 87.8%

Scale

$299

Cap a run: $0.02787
Margin: 80.1%

How it earns: DevRel and docs teams pay a monthly tier, Free, Team $49 or Scale $299. Plan, no paying customers yet. BUSINESS_PLAN.md

C1
Costs measured from 88 traces
C2
Caps written to the CLI, 2026-10-11
C3
Receipt: $0.0453, 96.4% of Team cap
C4
51 runs' cost against the 3 caps
C5
Cap alerts fired 3 times in 2 days

Show details

neatlogs

An AI judge named the right failed step in 46 of 52 runs. The audit judge averaged 4.87 of 5 over 15 audits. Scores

Investigate traced invented planner steps to the prompt, and commit 3ca6f11 fixed them. Investigate result

N1, N2 and N5 to N7 come from the dashboard on Oct 10. The repo does not record their trace IDs.

cfo.ai

Scenario, from $5,000	Month 18 cash	Runway
Base, $2,069 a month revenue	$16,685	No limit
Every run at $0.05286	$12,503	No limit
Scale at $529	$21,146	No limit
Old $0.50 Scale cap	-$12,212	Below $0, month 14

Average run $0.01782 over 88 traces. Each cap keeps a 70% margin. Revenue per run: Team $0.1633, Scale $0.0997. No customer has paid yet. The PR 15 run stopped at the Free cap of the time, $0.0108. PR 15 cfo-model.json

Entire output from the repo

A checkpoint on the planner fix commit3ca6f11

$ git log -1 --format="%h %s%n%(trailers:key=Entire-Checkpoint)" 3ca6f11
3ca6f11 Planner fix from neatlogs Investigate: copy commands only, guard against invented steps
Entire-Checkpoint: 01M4JGJ6NTE7YNB9Z0YW5FYVHN

entire-graph output, last line, trimmedgraph.ndjson

$ tail -1 evidence/graph.ndjson
{"record_type": "summary", "stats": {"files": 102, "symbols": 1587, "relations": 4644, "completeness_level": "ok"}}

Of the 41 commits without a checkpoint, 4 came before the first session. On many of the 37 site commits the Entire hook hung and was stopped. Trail reviews found 3 real bugs, all fixed. Findings. The Entire judge plugin scored 3.93 of 5, integrity 5.0, and flags thin commit-to-session coverage. jury-selftest.txt

20% DemoHow clearly the demo video explains the agent and your use of neatlogs, Entire and cfo.ai.

The 2:59 Demo Gives Each Tool Its Own Chapter

2:59

Each tool: architecture, implementation, runtime, reasoning, result. Times

Doc-audit has its own film: Find the Docs Your Code Left Behind, 1:56.

Show details

Part	neatlogs	Entire	cfo.ai
Architecture	1:10One run is one trace.	1:57A checkpoint per commit.	2:22Costs come from traces.
Implementation	1:16One init call.	2:00One command, before commit 1.	2:25Ari built 6 Scenarios.
Runtime	1:18A budget check before recovery.	2:03The fix session, prompt by prompt.	2:31Every Scale run at $0.50.
Reasoning	1:2552.8x incident, cause at 66%.	2:08Trails found 3 real bugs.	2:34Cash runs out, month 14.
Result	1:41Commit 2de852f fixed it.	2:1368 checkpoints on GitHub.	2:37A 2.8 cent cap stopped a run.

Chapters are the video's own: 0:00 problem, 0:12 dbt guide, 0:36 Docs Bisect, 0:57 the 4 PRs, 1:07 neatlogs, 1:51 Entire, 2:19 cfo.ai, 2:43 its own README. Times in the table come from the captions.

15% PublicHow you shared your project with other people while you built it.

11 Posts Tagged #neatHack Show the Build as It Happened

12:30 kickoff X X X X X X X X in X X Sat 10:00Sat 18:00Sun 00:00Sun 08:00 IST

10 on X, 1 on LinkedIn. Green: after kickoff. Full log

Show all 11 posts

Sat 10:45my neatlogs hackathon project has already caught doc issuesX
Sat 11:00built this weekend at #neatHackX
Sat 12:32#neatHack kickoff, 12:30 ISTX
Sat 13:14#neatHack, day one. all three required tools are upX
Sat 13:26your quickstart is the only part of your productX
Sat 15:15a quickstart breaks the way an agent breaksX
Sat 18:15send me your quickstart and FirstRun runs itX
Sat 18:234 docs PRs opened on day one of #neatHackX
Sat 19:30Never let an AI agent fix every failure on its ownLinkedIn
Sat 21:15an agent that heals every failure on its own driftsX
Sun 07:32a maintainer who gets a PR from an agent asks one questionX

10% UsefulWhether the agent solves a problem that people have, in a new way.

FirstRun Found 4 Real Docs Bugs and Sent 4 Upstream PRs

P1
dbt #10149: jafgen --years is gone
P2
dbt #10150: guide needs Python 3.12
P3
Entire #2724: entire not on PATH
P4
AccuKnox #671: latest is 0.9.58, not 0.9.0

$0model calls to find the bad release 76checkable receipts 7repos audited

Show details

All 4 PRs are open. Each came from a run verified in a fresh container.

firstrun receipt --check fails when a receipt or its evidence changed. A run receipt An audit receipt Receipts re-issued Oct 11 from saved run state.

Each Doc-audit board maps who works on what, from 14 to 247 people. dbt-core shows 15 stale pages and 72 features with no docs. dbt-core board

Older audits show an estimated cost, $0.0001 to $0.0055. How it is measured

Five Things Broke While Building, and Each One Has a Number

206 sGit 2.33 hung the Entire hook.
3traces from one run, now one.
50empty traces per test run, now 0.
39%higher cost count than neatlogs.
2Dagster gaps already documented, no PR.

Full list known-limits.md

Verify the Repo From Your Own Terminal

First commits and their dates

$ git clone https://github.com/HighnessAtharva/firstrun && cd firstrun
$ git log --reverse --date=iso --format="%h %ad %s" | head -3

First line: 96541a8 2026-10-10 12:34:51 +0530

Commits with a checkpoint

$ git log --no-merges --format="%(trailers:key=Entire-Checkpoint,valueonly)" | grep -c .
$ git rev-list --no-merges --count HEAD

91 and 132 on main up to f5d73eb, 2026-10-11. Both grow.

Offline tests384 passed

$ pip install -e ".[dev]" && pytest -q

No keys needed. 65 live tests skip. Actions

Run it on any quickstart

$ pip install git+https://github.com/HighnessAtharva/firstrun
$ firstrun https://docs.example.com/quickstart

Python 3.11, Git, Docker, a Claude token. README

neatHack 2026 · For judgesCheck Every FirstRun Claim in 60 Seconds

All 7 Official Evidence Rows Have Proof You Can Open

1The agent improved

2You used neatlogs

3You used Entire

4You used cfo.ai

5The agent recovers from failures

6You built in public

7You wrote the code during the hackathon

The Agent Runs a Quickstart Start to Finish, and v2 Beat v1

Same 4 quickstarts, v1 against v2

Recovery in trace 5520c5cf

Every part of a run points to code

All six numbers

neatlogs Improved the Agent, Entire Recorded the Build, cfo.ai Planned the Business

neatlogs

Entire

cfo.ai

Free

Team

Scale

neatlogs

cfo.ai

Entire output from the repo

The 2:59 Demo Gives Each Tool Its Own Chapter

11 Posts Tagged #neatHack Show the Build as It Happened

FirstRun Found 4 Real Docs Bugs and Sent 4 Upstream PRs

Five Things Broke While Building, and Each One Has a Number

Verify the Repo From Your Own Terminal