The logo, the colours, the one-liners, and the complete register of numbers we publish —
each with the source it was counted or corroborated from, and a dagger on the two we did
not count ourselves. Then the claims that failed verification, which is the more useful
half.
01 — The register, safe to quote
Every number we publish, and the source it came out of.
This is the whole list. If a figure about Arhas or one of its products is not in this
table, we did not publish it, and you should ask whoever did where they got it. Two rows
carry a dagger, because two of these figures are the product's own count rather than ours,
and the difference is the whole point of the page.
LabOS labos-website/src/data/site.ts:68-77 — the PLANS constant
403
Returned before a technician can approve their own result
LabOS api/results/route.ts:265-277 — rejected before any write
† The product's own published count, corroborated
by the module existing in the shipping source, but not independently recounted by us. Two
figures carry it — MedOS's 690 templates and its nine regulations — and they may never be
described as audited, verified or independently counted, here or anywhere else. They are
listed on their own in docs/CLAIMS.md §2.5.
Counted or corroborated from the shipping source on 14 July 2026 and held in src/data/products.ts, which is the same file this page renders from. A number
cannot appear on this site without passing through it, and it cannot pass through it
without a citation. The one-liners in section 03 are cleared for the same use.
02 — Tested and rejected, do not quote
The claims that failed verification, and why each one failed.
Every line below is a sentence somebody in this industry is printing today. Several are
sentences our own product sites printed until recently. They are quoted here for one
reason: to state plainly that they are not true of us, so that nobody has to take our
silence on trust. If you are writing about Arhas and a source hands you one of these,
the source is wrong.
The claimWhy it failedWhat we say instead
Rejected claim: Any customer count, testimonial, logo wall or star rating
Why it failed: There is no customer on record anywhere in our systems. The testimonials currently on the MedOS and LabOS sites are invented people — the LabOS site prints "representative examples shown" beneath its own. An invented person is not social proof. It is a lie with a headshot.
What we say instead: Nothing at all. When a real customer agrees to be named, they will be named, and not one hour earlier.
Rejected claim: ISO 27001 · SOC 2 · any external security certification
Why it failed: We hold none of them. No certificate exists, no audit has been run, and no badge goes on a website to represent an audit that has not happened.
What we say instead: We hold no external security certification, and the trust page says so in those words rather than routing around it.
Rejected claim: Independently security-reviewed · a third-party penetration test
Why it failed: No third party has reviewed this code. The only artefact is an internal security and compliance pass, which is a different thing and must not be dressed as the other one.
What we say instead: Hardened in an internal security pass. No external review has been commissioned. When one is, we will publish who did it and what they found.
Why it failed: MedOS runs against the ABDM sandbox. The NHA application had not been submitted as of May 2026, and no patient data has ever flowed through an ABDM endpoint. The certificate does not exist, so the word cannot be used.
What we say instead: ABDM-ready architecture, sandbox-validated. HIP certification in progress.
Rejected claim: DISHA compliant
Why it failed: DISHA was never enacted. It is a draft bill. Nothing can comply with a law that does not exist, and a vendor who claims to have complied with it is telling you they never read it.
What we say instead: Nothing. It is named here only in order to say that it is not law.
Rejected claim: NABL-certified · NABH-accredited · ISO 15189 certified
Why it failed: NABL accredits laboratories. NABH accredits hospitals. Neither body accredits software, so no software holds either accreditation — ours very much included.
What we say instead: LabOS is built for NABL: the hash-chained audit trail, the two-tier release and the report format are written for the standard the lab is assessed against. The lab is accredited. The software is not.
Rejected claim: HIPAA compliant · HIPAA certified
Why it failed: There is no such thing as a HIPAA certification, and HIPAA is United States law which does not bind an Indian clinic in any case. It appears on Indian health software because it sounds expensive.
What we say instead: HIPAA-aligned safeguards where they are useful. The law that actually governs an Indian clinic is Indian law.
Rejected claim: 99.98% uptime · any uptime or SLA percentage
Why it failed: We run no public status page and export no monitoring data, so there is nothing you could check the number against. A figure you cannot check is a figure we should not print.
What we say instead: Nothing, until there is a status page. When one exists, the number and the page that proves it will arrive on the same day.
Rejected claim: Sub-40 ms latency · <38 ms edge response
Why it failed: The Mumbai placement hint is real and verifiable in the deployment config. The latency figure was never measured, on any network, at any time. One does not imply the other.
What we say instead: Compute runs with a Mumbai placement hint. We publish no latency number, because we have not taken the measurement.
Rejected claim: +300% FPS · up to 40% faster · 4.2M+ results processed
Why it failed: The first two are performance promises about a machine the software has never seen. The third implies a production scale and a customer base that do not exist. All three were on a site of ours; none of them is on this one.
What we say instead: BRUTAL measures your machine, prints what it measured, and shows you the cases where it cannot help you at all.
This list is enforced, not aspirational. scripts/verify-claims.mjs greps the
entire source tree for these strings before astro build runs, and the build fails
if one of them appears anywhere it is not being explicitly rejected. The only way a banned
claim reaches this website is if somebody deletes the check.
Each one is checked against the same register as the numbers. Quote them whole — a number
paraphrased out of a sentence loses the citation that made it publishable.
The company
Arhas
Software you can check. We publish the price, the mechanism, and the limits. Then we ship the code that enforces them.
Arhas builds software for buyers who cannot audit a vendor, and makes the claim auditable
anyway. It is the parent of BRUTAL Optimizer, a Windows optimizer; MedOS, a hospital
management system for Indian clinics; and LabOS, a cloud laboratory information system for
Indian diagnostic labs. Every guarantee the software makes is enforced in the code path
rather than the manual, and every number it publishes carries the file it was counted from.
The legal entity name is
[legal name pending verification]
. Until it is confirmed against the certificate of incorporation, write Arhas with no legal suffix. Do not write Arhas Inc, Arhas Pvt Ltd, or any variant of either.
Windows performance tuning
BRUTAL Optimizer
A Windows optimizer that refuses to fake a number — 22 modules, 17 of them free, 661 named tweaks, each one reversible.
BRUTAL Optimizer is a Windows 10 and 11 tuning tool with 22 optimization modules — 17 of them free — and 661 named tweaks, each one a real Windows setting with its own undo. It ships no kernel driver and no game hooks, and its Smart Route relay engages only when it measurably beats your direct connection; otherwise it tells you to stay direct.
src/data/products.ts — summary and pitch
Hospital management for India
MedOS
A cloud hospital management system for Indian clinics, from ₹699 a month — with the price published and no quote form.
MedOS is a cloud hospital management system for Indian clinics and hospitals — appointments, records, lab, pharmacy, GST billing and insurance on one platform, from ₹699 a month with the price on the page and no quote form. It ships 690 consultation templates across 36 specialties, and it enforces Indian healthcare regulation inside the workflow rather than in a compliance PDF.
A cloud LIMS for Indian diagnostic labs, from ₹999 a month — accession to signed report, with 171 analyzers supported out of the box.
LabOS is a cloud laboratory information system for Indian diagnostic labs, from ₹999 a month — accession through to signed report, with 171 preset analyzer profiles across 43 manufacturers. Two-tier verification is enforced in the API, so the person who approves a result can never be the person who entered it.
src/data/products.ts — summary and pitch
04 — Logo and wordmark
A graduated scale that reads as the letter A.
Five ticks rising to an apex over a baseline rule. Measurement is the one idea all three
products share, so the mark is the visual form of it. It is not an illustration and it does
not want a gradient.
The mark is achromatic: ink on paper, white on ink. Do not recolour it, and do not tint it with a product accent — the parent hosts three colours and owns none of them.
No gradient, no rounded-square container, no drop shadow. Both child products use that formula. The parent must not, or it becomes a fourth sibling.
Clear space on all four sides is the height of one tick of the mark. Below 16px, use the mark alone and drop the wordmark.
The wordmark is "Arhas". Uppercase inside the lockup, sentence case in a sentence. Never ARHAS in running text, never Arhas™.
Product names are BRUTAL Optimizer, MedOS and LabOS. BRUTAL is always uppercase. Not Brutal, not MEDOS, not Lab OS.
05 — Colour
The parent is achromatic. The colour belongs to the products.
Every hue is already spoken for by a child product or carries a semantic meaning in all
three design systems, so Arhas takes none of them. Chroma enters an Arhas layout in exactly
three places: a 2px top rule on a product card, a small accent tick, and a hover wash.
Never a gradient, never a blurred blob, never a hero background.
Ink and paper
Ink 900Headlines, the footer ground, the primary button. 18.94:1 on paper, 19.58:1 on card white.#0A0C0F
Ink 500Body, muted text, citations and eyebrows. 6.03:1 on paper — the lightest ink allowed to carry prose on a light ground.#5A616D
Ink 4003.10:1 on paper. It fails AA and must never carry text on a light ground. It is for borders and icons — and for text on ink, where the polarity flips and it measures 6.10:1.#8A909B
Ink 200Borders, hairline rules, the graduated tick rule. 1.21:1 on paper, so never text of any size.#E3E6EA
PaperThe page itself. A cool white. Panels sit on --card #FFFFFF, and every ratio here is quoted against paper, the darker of the two.#FBFBFC
The three hosted accents
BRUTAL Optimizer4.30:1 on paper, 4.45:1 on pure white — under the 4.5:1 floor on the ground it actually sits on. Fills, 2px rules, icons and 18px-and-up bold text only. For green prose use --brutal-text #0A6E31: 6.18:1 on paper.#0C8A3D
MedOS5.48:1 on paper, 5.67:1 on card white. Passes AA for body text on both.#3B5BDB
LabOS5.51:1 on paper, 5.70:1 on card white. Passes AA for body text on both.#7C3AED
src/styles/tokens.css — the file the site itself renders from
If a fact you want is not on this page, it is not one we can stand behind.
Ask anyway. If we can source it, it gets added here where everyone can see it — not sent to
you in an email where nobody else can check it.
