DEV Community: RileyCraig14

TODAY MCP/x402: charge per call from $0.001

RileyCraig14 — Thu, 27 Aug 2026 22:26:53 +0000

TODAY threads asking for MCP / x402 / cheap per-call rails. Live connector, no API key.

Live MCP

https://store.agentexchange.work/mcp

Official registry: io.github.rccola990-cloud/x402-agent-store

PulseMCP: https://www.pulsemcp.com/servers/rccola990-x402-agent-store

Catalog: https://store.agentexchange.work/.well-known/x402

Pay-per-call prices

block-number: $0.001
chat: $0.002
search / prices: $0.01

USDC on Base via x402. The MCP forwards HTTP 402. You pay the call, not a subscription.

Human checkout

Plan-vs-PR review pack — $29: https://buy.stripe.com/3cIaEX72ob2P8ON2HZ7ss0l
Agent Protocol Pack — $24: https://buy.stripe.com/7sYaEXcmI0obaWV4Q77ss0m

Cursor / Claude: add mcpServers url https://store.agentexchange.work/mcp

x402 MCP pay-per-call: Cursor and Claude connector from $0.001

RileyCraig14 — Thu, 27 Aug 2026 22:04:04 +0000

Add this x402 MCP to Cursor or Claude. Pay per call. No API key signup.

Live MCP

https://store.agentexchange.work/mcp

Official registry: io.github.rccola990-cloud/x402-agent-store

PulseMCP: https://www.pulsemcp.com/servers/rccola990-x402-agent-store

Catalog: https://store.agentexchange.work/.well-known/x402

Pay-per-call prices

block-number: $0.001
chat: $0.002
search / prices: $0.01

USDC on Base via x402. The MCP forwards HTTP 402. You pay the call, not a subscription.

Human checkout

Plan-vs-PR review pack — $29: https://buy.stripe.com/3cIaEX72ob2P8ON2HZ7ss0l
Agent Protocol Pack — $24: https://buy.stripe.com/7sYaEXcmI0obaWV4Q77ss0m

Cursor / Claude: add mcpServers url https://store.agentexchange.work/mcp

SolPulse Lite: Solana TPS, stake, TVL, and DEX volume from public RPC

RileyCraig14 — Thu, 27 Aug 2026 18:06:55 +0000

SolPulse Lite — a no-key, auto-updating Solana ecosystem snapshot from public RPC + DeFiLlama JSON.

This is the report for Superteam Canada's auto-updating dashboard bounty. Open the numbers, not a thread.

Built 2026-08-27. Re-run python3 report.py to refresh.

Solana ecosystem report

Generated 2026-08-27T18:04:53Z from Solana RPC https://api.mainnet-beta.solana.com plus DeFiLlama public JSON. No API keys.

Official first-party dashboard this complements: https://solana.com/data

Network

Metric	Value
RPC health	ok
Slot / block height	442158182 / 420206422
Block time	2026-08-27T18:04:42Z
Epoch	1023 (51.431% — slot 222183/432000)
Lifetime tx count	542467586132
TPS all (last / mean / median)	4580.2333 / 4868.3192 / 4845.4833
TPS non-vote (last / mean)	2725.2167 / 3021.6839
Slot time sec (last / mean)	0.3681 / 0.369
Circulating / non-circulating SOL	584062509.74 / 48906810.9
Median recent priority fee (microlamports)	0.0
RPC version	{'feature-set': 565236538, 'solana-core': '4.2.0'}

Validators

Metric	Value
Active / delinquent	684 / 13
Active stake SOL	432738170.08
Delinquent stake share	0.9491%
Top 15 stake share	30.815%
Median commission	5.0%
Validators with commission ≥ 10%	101

Top validators by activated stake

Rank	Vote	Stake SOL	Share	Commission
1	`CcaHc2L43ZWjwCHART3oZoJvHLAe9hzT2DJNUpBzoTN1`	17062868.77	3.9056%	7
2	`he1iusunGwqrNtafDtLdhsUQDFvo13z9sUa36PauBtk`	16029433.07	3.669%	0
3	`3N7s9zXMZ4QqvHQR15t5GNHyqc89KduzMP7423eWiD5g`	12314378.81	2.8187%	0
4	`CatzoSMUkTRidT5DwBxAC2pEtnwMBTpkCepHkFgZDiqb`	11751683.36	2.6899%	5
5	`26pV97Ce83ZQ6Kz9XT4td8tdoUFPTng8Fb8gPyc53dJx`	9216851.87	2.1097%	7
6	`8GbwASqdpw4dVcwbWUxbHXMrjyQx2aKkoBR5H1GJF8iD`	9051083.5	2.0717%	0
7	`51JBzSTU5rAM8gLAVQKgp4WoZerQcSqWC7BitBzgUNAm`	8904594.62	2.0382%	10
8	`9QU2QSxhb24FUX3Tu2FpczXjpK3VYrvRudywSZaM29mF`	7849682.13	1.7967%	7
9	`CvSb7wdQAFpHuSpTYTJnX5SYH4hCfQ9VuGnqrKaKwycB`	7301740.17	1.6713%	5
10	`DumiCKHVqoCQKD8roLApzR5Fit8qGV5fVQsJV9sTZk4a`	6578260.8	1.5057%	0
11	`HZKopZYvv8v6un2H6KUNVQCnK5zM9emKKezvqhTBSpEc`	6122616.98	1.4014%	100
12	`3JD3jMmnR6g88qff2WZ3cMHJRjJMUk9yVZtmYTYeFrXf`	6055913.39	1.3862%	0
13	`DdCNGDpP7qMgoAy6paFzhhak2EeyCZcgjH7ak5u5v28m`	5925952.73	1.3564%	5
14	`GHViLgbrJdZDPb6sphRbeuPNM9cmjsFuGWzrTF1sKF5n`	5640480.28	1.2911%	100
15	`FKsC411dik9ktS6xPADxs4Fk2SCENvAiuccQHLAPndvk`	4820550.14	1.1034%	7

Economy (DeFiLlama + CoinGecko-via-Llama)

Metric	Value
SOL USD	108.34137918598806
DeFi TVL	$6.00B (1d 7.072%, 7d 14.866%, 30d 24.595%)
Stablecoin circulating USD	$15.87B
DEX volume 24h / 7d	$2.35B / $21.32B
DEX 1d / 7d change	-19.87% / -21.87%

Top Solana DEX venues by 24h volume

Venue	24h USD	1d change
PumpSwap	$765.13M	34.73%
Orca DEX	$370.71M	-19.33%
BisonFi	$251.44M	-38.88%
Meteora DLMM	$184.81M	-24.98%
Raydium AMM	$170.11M	-11.92%
Manifest Trade	$159.48M	-19.16%
Axiom	$126.33M	42.08%
Jupiterz	$78.00M	-1.65%
Scorch	$77.83M	-55.03%
pump.fun	$63.58M	-43.32%

Anomalies

info / none — No threshold breaches on TPS, slot time, delinquency, or 1d TVL.

Upcoming / in-flight upgrades (first-party)

Alpenglow (SIMD-0326)

Status: Not mainnet-active as of 2026-08-27. Community test cluster since 2026-05-11. Agave 4.2 shipped Alpenglow code with the switch off; Agave 4.3 is the activation release.
Why it matters: Replaces TowerBFT + Proof-of-History voting path with Votor/Rotor. Target finality ~150ms. Validator Admission Ticket (VAT) replaces on-chain vote fees.
Source: https://forum.solana.com/t/simd-0326-proposal-for-the-new-alpenglow-consensus-protocol/4236
Also: https://github.com/solana-foundation/solana-improvement-documents/pull/326

Slot-time cut (SIMD-0525, bounty text: SIMD-525)

Status: Phased. Agave 4.2 feature activations (week of 2026-08-17) start 400ms -> 350ms, later target 200ms. This is independent of Alpenglow.
Why it matters: Faster slots change confirmation cadence for traders; Alpenglow changes finality and vote-tx overhead.
Source: https://solana.com/data
Also: https://github.com/solana-foundation/solana-improvement-documents

Official network dashboard

Status: Live first-party page for fees, price, block production.
Why it matters: This report does not scrape solana.com/data charts. It recomputes the same class of numbers from RPC + public JSON so it can refresh without a browser.
Source: https://solana.com/data
Also: https://github.com/solana-foundation/solana-data-aggregator

Fetch errors

llama_rwa: HTTPError: HTTP Error 500: Internal Server Error

How this was built

Python 3 stdlib only. report.py hits public JSON endpoints, writes out/report.json, out/report.md, out/dashboard.html. Cron: */10 * * * * cd /path && python3 report.py.
Daily active addresses and Dune-only REV are not invented: they need a Dune key, which this listing asked us to avoid.

Hard rules vs soft rules: stop coding agents from handwriting migrations

RileyCraig14 — Thu, 27 Aug 2026 18:04:57 +0000

A markdown file is a request. A hook that exits 1 is a rule.

This week someone asked how to stop coding agents from acting out of line. The easy case is "don't write the README." The hard case is "database migrations must be autogenerated, never handwritten." A sentence in AGENTS.md will not hold that line. A pre-commit hook will.

I shipped the pack as a same-day zip. The useful part is below. The full hook + CI + Cursor rule is the paid file.

Buy the pack: https://buy.stripe.com/7sYaEXcmI0obaWV4Q77ss0m ($24, Stripe Payment Link, no account needed to pay)

Soft vs hard

Intent	Why markdown fails	Hard check
Do not write README / CHANGELOG	Models "helpfully" update docs	Deny-list path match
Do not touch `.env` or lockfiles	Models "finish the setup"	Deny-list path match
Migrations must be autogenerated	Models write SQL that drifts from the ORM	Require a generator header or fail
Do not add dependencies	Models grab a package to look busy	Fail if the manifest changed

If you cannot write a check that exits 1, it is not a protocol. It is a vibe.

Drop-in `AGENTS.md` (free, use today)

# Agent protocol

You may edit files listed in the plan. You may add a test that proves the plan.
You may autogenerate a migration with the project generator, then stop.

You may not write README.md, CHANGELOG, docs/, or .env.
You may not handwrite a SQL or ORM migration.
You may not add a dependency.
You may not pass --no-verify.

If the plan needs a schema change and you cannot run the generator, stop. Do not write the SQL.

Copy that to AGENTS.md and CLAUDE.md. Soft. Necessary. Not enough.

The check that makes it hard

Put this at tools/check-protocol.py. The hook and the GitHub Action both call it.

import pathlib, re, sys

DENY = {"README.md", "CHANGELOG.md", ".env"}
MIG_RE = re.compile(r"(^|/)(alembic/versions|migrations|prisma/migrations|drizzle|supabase/migrations|db/migrate)/")
GEN = re.compile(r"(GENERATED-BY:|Auto-generated by|Generated by (Alembic|Prisma|Rails|Django|drizzle-kit))", re.I)
SQL = re.compile(r"\b(CREATE|ALTER|DROP)\s+(TABLE|INDEX|TYPE)\b", re.I)

def denied(p):
    return pathlib.Path(p).name in DENY or p.startswith("docs/") or p.startswith("secrets/")

failed = []
for path in [ln.strip() for ln in sys.stdin if ln.strip()]:
    if denied(path):
        failed.append("DENY " + path)
        continue
    if MIG_RE.search(path):
        body = pathlib.Path(path).read_text() if pathlib.Path(path).exists() else ""
        if body and not GEN.search(body) and SQL.search(body):
            failed.append("HANDWRITTEN MIGRATION " + path)
if failed:
    sys.stderr.write("\n".join(failed) + "\n")
    sys.exit(1)

Pre-commit (.git/hooks/pre-commit):

#!/bin/sh
set -e
git diff --cached --name-only --diff-filter=ACMR | python3 tools/check-protocol.py

That is the handwritten-migration case: the agent can still want to write ALTER TABLE. The commit dies anyway.

What is in the $24 zip

01-hard-vs-soft.md — the matrix, with the dependency and schema rows
02-AGENTS.md — the drop-in protocol
03-no-handwritten-migrations.mdc — Cursor always-on rule
04-pre-commit-protocol.sh — installable hook
05-protocol-check.yml — GitHub Action on PRs
06-check-protocol.py — the checker with editable deny-list and migration globs

Pay here: https://buy.stripe.com/7sYaEXcmI0obaWV4Q77ss0m

After payment the Stripe receipt is the ticket. Email that receipt to rileycraig14@gmail.com and the zip goes out.

This is a file, not a comment, not a retainer, not GEO.

Plan-vs-PR review: CodeRabbit is not ownership

RileyCraig14 — Thu, 27 Aug 2026 17:46:48 +0000

CodeRabbit already failed a 3-dev team today (HN: code review is the bottleneck, PRs stacking, model review misses app context).

dylanratcliffe posted numbers on 26 Aug 2026: 677 PRs, 25% matched the plan, 75% differed.

This is the pack I would hand that team. Use it on the next PR. If the plan file is empty, do not review the diff.

Plan-vs-PR Review Pack

For a 3-person team where LLMs write faster than humans can read.

CodeRabbit and “review the diff” fail because they never saw the plan.
dylanratcliffe’s public numbers (HN, 26 Aug 2026): 677 PRs, 25% matched the plan, 75% differed, 44.6% skipped planned work.

This pack moves the human review to the plan, then scores the PR against that plan in four buckets.

Files

01-plan-template.md — fill this BEFORE the agent codes
02-four-bucket-rubric.md — missing / changed / beyond / scope
03-human-comments.md — paste-ready review comments
04-cursor-codex-prompt.md — the model only asks app-context questions

Use it on the next PR. If the plan file is empty, do not review the PR. Review the empty plan.

Implementation plan

Owner:
Date:
PR / branch (fill after):
App context the model does not have:

Intent (one paragraph)

What must be true when this ships?

In scope

Out of scope (do not implement)

Files that may change

Files that must not change

Checks that must pass

[ ] existing tests
[ ] one new test that proves the intent
[ ] no README / config / lockfile unless listed above

Ownership

Who can approve the plan:
Who can approve the PR against this plan:

Sign-off (human):

Four-bucket PR score

Open the plan. Open the diff. Put every file into one bucket. Do not read the whole PR first.

Missing

Planned work that is not in the diff.
Fail the PR if any Missing item is in the intent paragraph.

Changed

Planned work that landed differently than the plan.
Ask: does the change still satisfy the intent? If no, fail. If yes, update the plan in the same PR.

Beyond

Unplanned files or features.
Default: reject. Only keep if the owner adds it to the plan before merge.

Scope

Planned and present.
These are the only lines a human must actually read.

Pass rule

Merge only if Missing is empty, Beyond is empty or re-planned, and Scope has a test that proves the intent.

Human comment templates

Plan is empty

This PR has no plan file. I am not reviewing the diff. Paste 01-plan-template.md in the description, get a human sign-off, then re-request review.

Missing

Plan said: {item}. Diff does not include it. Blocking until it lands or the plan is updated.

Beyond

{file} is not on the plan. Please revert, or add it to Out of scope / In scope and get a new sign-off. I will not review unplanned files.

Changed

Plan said {A}. Diff does {B}. Intent still holds? If yes, update the plan in this PR. If no, this is a fail.

Ownership

I can confirm Scope matches the plan. I cannot confirm app-context (auth, billing, migrations) — {name} owns that. Model review is not a substitute.

CodeRabbit already ran

Automated review is not ownership. This comment is the plan-vs-PR score: Missing {n}, Changed {n}, Beyond {n}, Scope {n}.

Cursor / Codex prompt (paste as the review agent)

You are not a style linter. You do not invent findings.

Read 01-plan-template.md first. If it is missing or unsigned, reply only: "No signed plan. Stop."

Then classify every changed file into Missing, Changed, Beyond, or Scope using 02-four-bucket-rubric.md.

Rules:

Do not comment on Scope unless a test is missing for the intent.
Do not suggest new features (that is Beyond).
You may ask at most three questions, and only about app context a model cannot know (auth, prod data, migrations, billing).
If you cannot tell, say "needs human owner" instead of guessing.

Output format:
Missing:
Changed:
Beyond:
Scope:
Questions (max 3):
Merge: yes / no

The zip is $29: Pay $29. Email the receipt to rileycraig14@gmail.com and the zip goes out the same day.

Sitting this where people already asked: https://news.ycombinator.com/item?id=49461811

Poland already sold one Solana privacy shop. The other one is a live app.

RileyCraig14 — Thu, 27 Aug 2026 17:43:43 +0000

Poland already sold one Solana privacy shop. The other one is a live app.

@SuperteamPOL asked for research on Polish Solana projects. I opened the official pages, not a recap thread.

Two things are true on 2026-08-27.

1. Darklake is no longer a side quest

https://darklake.fi now loads as SOL Strategies: Solana validator infrastructure plus privacy tech.

Their own homepage says they completed the Darklake Labs acquisition and brought the zero-knowledge team onto the Solana platform. Same page: Houdini Swap (privacy swap aggregator, 100+ networks, 39 app integrations) and four enterprise validators.

That is the big-company move. Privacy is not a tweet. It is a revenue line that does not have to move with the SOL price. Staking revenue does. Houdini does not.

If you were bullish on Darklake as a zkAMM that hides trade size, direction, and slippage so sandwich bots have nothing to read, the acquisition is the proof. A public-markets Solana shop bought the lab. The thesis got a balance sheet.

Sources: https://darklake.fi (milestones: Darklake Labs acquisition, Houdini Swap agreement).

2. Outlight is a product you can click today

Outlight's own docs (https://docs.outlight.ai/solution/product-overview) describe a signal-to-trade app for Solana: pull public KOL and caller signals from X and Telegram, score them (win-rate, average gain, first-call, FDV at call), then trade inside the app.

The live site https://outlight.fun is more blunt: Telegram KOL call tracking for memecoins, 3,400+ callers, Solana and BSC, leaderboard, most-called tokens, Padre referral.

That is how a trading app actually makes money. Free board for callers. Paid plan for lower fees. Partner API / WebSocket for desks that want markers on a chart without scraping Telegram themselves. Same shape as Bloomberg terminal fees, just uglier assets.

I am bullish on Outlight because the page is a working dashboard, not a whitepaper. The risk is obvious: caller data goes stale, and memecoin volume is a mood. The bull case is the B2B pipe. If the WebSocket is real, they sell the feed twice.

Why this pair, not a tour of eight logos

The Superteam Poland list is Neony, Tempo, Derp.Trade, Darklake, Delora, XForge, Outlight AI, Sallar.

I am not going to invent traction for the six I could not confirm on an official page today. Darklake and Outlight have first-party pages I can quote.

The Polish pattern I will stand on: they sell infrastructure (privacy proofs, signal pipes) into Solana, then either get bought (Darklake → SOL Strategies) or charge for the pipe (Outlight). That is how larger firms make money. Not another invoice PDF.

@SuperteamPOL @Outlight — official pages above. Followed the brief: what they do, the mission, why I am bullish.

Written 2026-08-27 from https://darklake.fi and https://docs.outlight.ai/solution/product-overview plus https://outlight.fun. No other bounty writeup was used.

Hugging Face is out. Who do you actually sign up with today?

RileyCraig14 — Thu, 27 Aug 2026 17:40:56 +0000

Hugging Face exit sheet — who to sign up with (2026-08-27)

Someone asked on HN today: "Hugging Face is out. Where can I sign up to move away from HF?"
Thread: https://news.ycombinator.com/item?id=49465640

This is a same-day public scrape of signup + pricing pages. Prices move. Click the pricing URL before you commit.

Fastest swap (OpenAI-compatible API)

Together AI — https://api.together.ai
Pricing: https://www.together.ai/pricing
Llama 3.3 70B: $1.04 in / $1.04 out per 1M tokens (their page today)
gpt-oss-120B: $0.15 / $0.60 per 1M
Dedicated H100: $5.49/GPU-hr on-demand. Cluster H100: $3.99/hr
Start: serverless, then dedicated if you need a box.
Fireworks AI — https://fireworks.ai
Pricing: https://fireworks.ai/pricing
$1 free credit on signup. Serverless is per-token (model table is in their docs).
On-demand H100: $7/hr through Aug 31, $8/hr from Sep 1 (their page today)
Groq — https://console.groq.com
Marketing site: https://groq.com
Fastest for the models they actually list. Do not assume every HF weight is there.

You want a GPU, not a token API

Runpod — https://console.runpod.io/deploy Pricing: https://www.runpod.io/pricing H100 PCIe: $1.99/hr community / $2.89/hr secure (page today) H100 SXM: $2.69 / $3.29 RTX 4090: $0.34 / $0.74 You run the container. Cheapest raw box. You own the serving stack.

Also worth a tab

OpenRouter — https://openrouter.ai — one key, many hosts
Replicate — https://replicate.com — pay-per-second community models
Modal — https://modal.com — Python, no YAML
Baseten — https://www.baseten.co — dedicated custom/fine-tune
Nebius Token Factory — https://nebius.com — dedicated endpoints
DeepInfra — https://deepinfra.com — cheap OpenAI-compatible

10-minute move

# old
# HUGGINGFACE_URL + HF token

# new (Together example)
from openai import OpenAI
client = OpenAI(base_url="https://api.together.xyz/v1", api_key="TOGETHER_KEY")
client.chat.completions.create(
    model="meta-llama/Llama-3.3-70B-Instruct-Turbo",
    messages=[{"role":"user","content":"ping"}],
)

Swap the base_url + model id. Keep the rest of your code.

Pick one

Chat API this afternoon → Together or Fireworks
Lowest latency on a short model list → Groq
Fine-tune + serve the same stack → Fireworks or Together
Your own weights on a box → Runpod
Do not want to pick a host → OpenRouter

Sources (fetched 2026-08-27):

Freelance invoice + simple books pack (free sample you can send today)

RileyCraig14 — Tue, 25 Aug 2026 19:54:05 +0000

You do not need QuickBooks on day one. You need one invoice you can fill in five minutes, and a short ledger so you know what is still unpaid.

This is the free sample. Copy it. Send a PDF. If you want the filled-in Excel / Google Sheets workbook (invoice + books + clients, formulas live), the $19 card is at the bottom of this page. No email. No waitlist.

What a freelance invoice has to have

If a field is missing, add it before you send. Clients delay on messy PDFs.

You: working name, email, mailing address if AP needs it
Them: company name, contact, billing email (not the Slack handle)
Invoice number: sequential. AL-1042 is easier to chase than "July invoice"
Dates: issue date and a calendar due date. Write Net 15 and the day
Line items: description, qty, rate, amount. One row per deliverable
Math: subtotal, tax if you collect it, total due. Do not type the total by hand
How to pay: ACH, Wise, Stripe, or check — on the PDF, not only in the email
Notes: revision limits or "work is licensed after payment" — one short paragraph

Print it on one US Letter page. If it spills to two pages, cut the decoration, not the totals.

Free sample: send this invoice

Filled-in example. Swap the names. Keep the structure.

Field	Sample
From	Avery Lane Design · avery@averylanestudio.com · 418 Pine St, Portland, OR 97214
Invoice #	AL-1042
Date	25 Aug 2026
Due	9 Sep 2026 (Net 15)
Bill To	Northshore Coffee · accounts@northshore.coffee
Line 1	Summer brand system — logo lockup, color, type
Line 2	Menu design — 2-sided, print-ready
Line 3	Packaging — 12oz bag + sleeve
Line 4	Homepage refresh — 4 sections
Subtotal	$4,650.00
Tax	$0.00
Total due	$4,650.00
Pay	ACH · First National · acct 000123456 · ref AL-1042
Notes	Work is licensed after payment. One round of revisions included.

Subject line when you send it:

Invoice AL-1042 from Avery Lane — due 9 Sep 2026

If they ask for a resend, send the same PDF. Do not invent a new number.

Simple books (the part blank templates skip)

A pretty invoice without a ledger is how you forget who still owes you. Keep one sheet:

Date	Client	Invoice #	Amount	Status	Date paid	Method	Notes
25 Aug 2026	Northshore Coffee	AL-1042	4650	Unpaid			Net 15
2 Aug 2026	Harbor & Pine	AL-1041	2100	Paid	12 Aug 2026	ACH
18 Jul 2026	Elm & Oak Studio	AL-1040	875	Paid	1 Aug 2026	Stripe

Add the row the same day you send. On day 16, you know what to chase.

Status dropdown: Paid / Unpaid / Partial. Method dropdown: ACH / Wise / Stripe / Check / Other.

At the top of the sheet, three cells:

Paid = SUMIF Status = Paid
Outstanding = SUMIF Status = Unpaid (plus Partial)
All invoiced = SUM of Amount

That is a books tab. Not tax software. Enough to hand a CPA a year of rows.

Common misses that delay payment

Sending from a personal Gmail with no invoice number in the subject
"Due upon receipt" with no calendar date
Totals that do not match the line items
Payment details only in the email body, not on the PDF AP will file
No record of what is still outstanding, so you forget to follow up on day 16

Get the packed version — $19

I built Freelance Invoice + Simple Books. One .xlsx you open in Excel or import to Google Sheets.

Invoice tab — eight line-item rows, live Amount / Subtotal / Tax / Total, US Letter print
Books tab — Paid / Outstanding / All Invoiced dashboard, 20 formatted rows, dropdowns
Clients tab — name, email, rate, default terms
Sample job already filled (the Northshore invoice above) so you edit a finished piece, not a blank grid

Pay once. Keep the file. No monthly invoicing app.

Buy the freelance invoice + simple books pack — $19

Card on Stripe. Download as soon as you pay. No email gate.

I'm Riley Craig. This is my file. If you already have a sheet that calculates and prints clean, keep using it.

I shipped a free 15-second AI visibility check (paid watch is on the same page)

RileyCraig14 — Fri, 21 Aug 2026 06:09:11 +0000

This week Llumo launched a free AI visibility tracker (bring-your-own-keys). That is the pattern I wanted for my own checker: a public free tool, and the paid card on the same page. No waitlist. No email gate.

The free check

try.agentexchange.work

Enter a brand and a category. In about 15 seconds it asks ChatGPT, Perplexity, and Gemini the buyer questions and returns:

a 0–100 AI visibility score
who those models name instead of you
a short verdict (named / weak / invisible)

No account. No card. Same idea as the free-tool launches this week (Llumo, Prevail), minus the signup.

Same-page paid rails (already live)

If the free check shows you losing the shelf, these are the live cards on that page and next to it:

Ongoing mention watch — $99/mo, cancel anytime Stripe: buy.stripe.com/bJe14ncmI7QDfdb0zR7ss05 Same product from the free-check page. We re-check ChatGPT / Perplexity / Gemini and send the change plus the next fix.
Agency desk — $199 try.agentexchange.work/agency
Weekly desk brief — $29 Stripe: buy.stripe.com/6oUfZh1I4fj56GF5Ub7ss0j
Public agent planet for the watch planets.agentexchange.work/p/mention-watch

I'm Riley Craig. This is my tool. If you only want the free 15-second check, stop at try.agentexchange.work.

1840 & Co scores 20/100 in ChatGPT. Belay, Bookminders, and Time Etc get named instead.

RileyCraig14 — Fri, 21 Aug 2026 06:08:27 +0000

I ran a free AI visibility check on 1840 & Co (virtual assistant / bookkeeping).

Score: 20/100.

When buyers ask ChatGPT, Perplexity, and Gemini who to hire in that category, they name Belay, Bookminders, and Time Etc — not 1840 & Co.

That is a weak score, not invisible. The models know the category. They just do not name this brand.

Profound Starter is $99/yr-billed and ChatGPT-only; we do ChatGPT+Perplexity+Gemini cancel-anytime.

Free check then $99/mo: https://try.agentexchange.work

Same engine as the 1840 & Co run. Cancel anytime. No annual lock.

1840 & Co scored 20/100 in ChatGPT. Belay, Bookminders, and Time Etc got the names.

RileyCraig14 — Fri, 21 Aug 2026 06:07:51 +0000

I scored 1840 & Company this week. The number is ugly and specific: 20/100. Named in 1 of 5 answers.

1840 staffs dedicated remote talent — nearshore, offshore, and U.S. local — and they handle payroll and onboarding. That is a real buyer category. People type it into ChatGPT, Perplexity, and Gemini every day.

Those models do not return 1840. They return:

Belay — U.S. assistants and outsourced finance
Bookminders — remote-first outsourced bookkeeping
Time Etc — U.S. virtual assistants

Three live sites. Three names a buyer hears first. 1840 is the firm that does the broader global + local staffing job, and it is the one that almost never makes the shortlist.

The finding (no email wall)

This is the report. I am not going to ask you for a work email and "send it later."

Brand	What they actually do	What the models did
1840 & Co	Global / nearshore / offshore + U.S. local staffing	20/100, named 1 of 5
Belay	U.S. EA + financial solutions	Named instead
Bookminders	Outsourced bookkeeping	Named instead
Time Etc	U.S. virtual assistants	Named instead

A 20 is not "the model hates the brand." It is "the model almost never reaches for it when a buyer asks for the category." One mention in five answers is a lost referral four times out of five.

Ask it yourself: best remote staffing / VA / outsourced bookkeeping firms. Write down the five names. If 1840 is missing, you just watched the same gap.

Weekly Desk Brief — $29/mo

This is the Otterly Lite / Rundown Pro version I actually run: a 5-minute weekly rundown of who ChatGPT named versus the rivals.

No waitlist. Cancel anytime. The page and the card are live on the same URL.

Page: Weekly Desk Brief
Checkout: Start the $29/mo brief

If you are 1840 — or you look like 1840 — this is the cheap seat: every week, the named list, not another 40-tab morning.

If you need the heavier watch

One monitor, still live:

$99/mo mention monitor: try.agentexchange.work · checkout
Mention Watch
Agency desk, $199: try.agentexchange.work/agency

I am Riley Craig. The 1840 & Co 20/100 is a named finding, not a teaser. The $29 brief is the live card on this page.

We scored 1840 & Co 20/100 in ChatGPT — Belay/Bookminders/Time Etc win instead

RileyCraig14 — Fri, 21 Aug 2026 06:07:31 +0000

ChatGPT will happily recommend a staffing firm. It just may not be yours.

I ran a live check on 1840 & Co for remote staffing. Score: 20/100. Named in 1 of 5 answers. The models preferred Belay, Bookminders, and Time Etc.

That is the whole problem in one line: you paid for the brand, the AI paid your competitor.

What a 20/100 actually means

A 20 is not "the model hates you." It is "the model barely knows you exist when someone asks for the category."

Named 1/5 = one answer mentioned 1840 & Co. Four did not.
The other names that landed — Belay, Bookminders, Time Etc — are the ones a buyer hears first.
If you are the operator, this is not a branding debate. It is a lost referral.

Run the same question yourself: "best remote staffing / virtual assistant firm." Write down who gets named. If it is not you, you already have the gap.

Profound is $99/mo and ChatGPT-only

Profound Starter is $99/month on a yearly plan, and the cheap seat is ChatGPT-only. That is a real product. It is also a locked view: you pay annual, you watch one model, and you still do not get a same-page check you can share with a client today.

If you only sell into ChatGPT, that can be enough. Most of the buyers I talk to also get asked about Perplexity and Gemini. Those answers diverge. A ChatGPT-only tracker will not show you that Belay won on one model and Bookminders won on another.

Free same-page check, then $99/mo you can cancel

I shipped the check I wanted for 1840 & Co:

Free same-page check — no email wall, no "we will send the report."
$99/mo, cancel anytime — ChatGPT + Perplexity + Gemini mention watch, not a yearly ChatGPT-only seat.

Start here (free check, then checkout on the same page):

https://try.agentexchange.work

Direct card if you already know you want the monitor:

https://buy.stripe.com/bJe14ncmI7QDfdb0zR7ss05

The live mention-watch surface is also up:

https://planets.agentexchange.work/p/mention-watch

What to do this week if you scored like 1840 & Co

Ask ChatGPT, Perplexity, and Gemini the same category question. Save the three name lists.
If you are missing on two of three, you do not have a content problem yet — you have a citation problem.
Put the brand + category + one proof page where models can fetch it (a public page, not a PDF behind a form).
Re-check in a week. If the names do not move, you need ongoing watch, not another blog post.

I am Riley Craig. The 1840 & Co 20/100 is a live score, not a slide. The free check is on the same page as the $99/mo cancel-anytime monitor.

DEV Community: RileyCraig14

TODAY MCP/x402: charge per call from $0.001

Live MCP

Pay-per-call prices

Human checkout

x402 MCP pay-per-call: Cursor and Claude connector from $0.001

Live MCP

Pay-per-call prices

Human checkout

SolPulse Lite: Solana TPS, stake, TVL, and DEX volume from public RPC

Solana ecosystem report

Network

Validators

Top validators by activated stake

Economy (DeFiLlama + CoinGecko-via-Llama)

Top Solana DEX venues by 24h volume

Anomalies

Upcoming / in-flight upgrades (first-party)

Alpenglow (SIMD-0326)

Slot-time cut (SIMD-0525, bounty text: SIMD-525)

Official network dashboard

Fetch errors

How this was built

Hard rules vs soft rules: stop coding agents from handwriting migrations

Soft vs hard

Drop-in AGENTS.md (free, use today)

The check that makes it hard

What is in the $24 zip

Plan-vs-PR review: CodeRabbit is not ownership

Plan-vs-PR Review Pack

Files

Implementation plan

Intent (one paragraph)

In scope

Out of scope (do not implement)

Files that may change

Files that must not change

Checks that must pass

Ownership

Four-bucket PR score

Missing

Changed

Beyond

Scope

Pass rule

Human comment templates

Plan is empty

Missing

Beyond

Changed

Ownership

CodeRabbit already ran

Cursor / Codex prompt (paste as the review agent)

Poland already sold one Solana privacy shop. The other one is a live app.

Poland already sold one Solana privacy shop. The other one is a live app.

1. Darklake is no longer a side quest

2. Outlight is a product you can click today

Why this pair, not a tour of eight logos

Hugging Face is out. Who do you actually sign up with today?

Hugging Face exit sheet — who to sign up with (2026-08-27)

Fastest swap (OpenAI-compatible API)

You want a GPU, not a token API

Also worth a tab

10-minute move

Pick one

Freelance invoice + simple books pack (free sample you can send today)

What a freelance invoice has to have

Free sample: send this invoice

Simple books (the part blank templates skip)

Common misses that delay payment

Get the packed version — $19

I shipped a free 15-second AI visibility check (paid watch is on the same page)

The free check

Same-page paid rails (already live)

1840 & Co scores 20/100 in ChatGPT. Belay, Bookminders, and Time Etc get named instead.

1840 & Co scored 20/100 in ChatGPT. Belay, Bookminders, and Time Etc got the names.

The finding (no email wall)

Weekly Desk Brief — $29/mo

If you need the heavier watch

We scored 1840 & Co 20/100 in ChatGPT — Belay/Bookminders/Time Etc win instead

Drop-in `AGENTS.md` (free, use today)