Investors - seed / pre-seed / strategic
The Pedulli Orchestrator (T15) is a lossless compression engine framed as a best-of-N racer — it races xz, zstd, brotli and your data's SRD math, then ships the smallest byte-exact result, with a +1-byte never-worse floor on its own output. Because it includes the standard codecs and keeps the smallest verified winner, it is never larger than the best standard codec (worst case +1 byte): it wins outright on structured data and ties the best codec on already-optimal or random data. Every benchmark we publish is Bitcoin-anchored for measurement priority.
This page is organised into three honest buckets. Everything in A is reproducible today; B is early-stage pipeline described generically; C is internal research under verification.
Bucket A — Verified results (reproducible today, with proof)
Each of these is backed by a script plus hash, a byte-exact roundtrip, a re-runnable artifact, or a Bitcoin-anchored manifest. Nothing here is a grand thesis — these are checkable engineering results.
An independent, reproducible benchmark: every input and output is SHA-256-hashed and roundtrip-verified byte-for-byte, with the result JSON committed by hash (re-confirmed: 26,487 bytes, matching the manifest). The methodology is Bitcoin-anchored (OpenTimestamps) for measurement priority. The harness even flags roundtrip failures rather than counting them as wins.
Our lossless compressor reconstructs the standard enwik8 / enwik9 corpora byte-for-byte (SHA-256-verified, full-pipeline-exact). To be explicit: this is not a Hutter-Prize-winning result — it ties xz on generic text and does not beat the cmix research record. We publish it only with that framing.
Byte-exact lossless with a verified +1-byte never-worse floor on Pedulli's own output. This is a guarantee about our output (worst case: input + 1 byte), not a claim of beating every codec. On already-optimal or random data we simply match that floor.
When a decision process has a real symmetry, solving the reduced (SRD) problem and lifting back reproduces the reference value function and policy bit-exactly (verified across the test configurations). A built-in check refuses the shortcut when the symmetry does not hold. This is value/policy preservation under symmetry — not a speedup claim. The script and result JSON are shippable so third parties can re-run and reproduce the result.
Checks supplied witnesses only: AES-GCM re-encryption match, p*q == N bigint check, SHA-256 roundtrip, supplied-clause SAT. It verifies what you give it. It does not break, crack, or solve RSA, AES, ECC or anything else.
Every benchmark manifest is anchored to the Bitcoin blockchain — independent block-height and timestamp evidence of measurement priority, verifiable in about a minute. The engine internals stay closed; the public surface is measured behaviour plus a protected runtime and exact verifier. No absolute-security claims.
Bucket A — measured wins, bounded never-worse
These are measured results on specific structured data classes. Because the Orchestrator races xz, zstd and brotli and keeps the smallest verified output, it is never larger than the best standard codec (worst case +1 byte): it wins outright on structured data and ties the best codec elsewhere. The full head-to-head table lives on the benchmarks page.
| Input class | Strong generic codec | Pedulli (best-of-N) | Result |
|---|---|---|---|
| MP4 master ~10 MB | xz-9 → 9,468,380 B | 9,430,108 B | -38,272 B |
| JSON ~31 KB | brotli-11 → 2,510 B | 1,265 B | -50% |
| Apache logs ~3.5 MB | xz-9 → 211 KB | 211 KB | selects xz — ties the best codec |
| Incompressible random 1 MB | xz +424 B / 7z +658 B | +1 B | never-worse floor |
| Long English text >100 KB | xz-9 → best generic | = xz-9 | selects xz — ties the best codec |
Figures are measured benchmark results with Bitcoin-anchored manifests. The +1-byte floor is a guarantee about Pedulli's own output, not a claim of beating every codec on every input.
Bucket B — Commercial pipeline (early-stage)
Pedulli is pre-revenue with no signed customers, no executed pilots, and no testimonials. We describe the pipeline honestly and generically — no company names, no logos — because nothing has been contracted.
Early conversations with an enterprise AI-platform provider around a verifier-gated retrieval / embedding-acceleration use case. No relationship exists yet: no contract, no pilot, no sent commitment.
Early conversations with a storage / compression-infrastructure provider about byte-exact lossless compression to reduce stored bytes per node and egress. Exploratory only — no contract, no pilot.
We are building a design-partner pipeline across storage / backup, media masters, and telemetry / log infrastructure. Reproducible demos are available under NDA. No named partners are presented as relationships.
An investor / accelerator scouting list (sourced from public information) exists for a planned pre-seed raise. No investor is committed; this is preparation, not traction.
Honest posture: early-stage, pre-revenue, building a design-partner pipeline; reproducible demos available under NDA. No customers, pilots, testimonials, or logos. Any prospect is referenced generically until a signed agreement or written testimonial exists.
Bucket B — business model
Personal use, no signup for the browser trial. Sub-1 MiB runs on our EU servers (deleted after the job). Free API tier with API key + token balance. Acquisition channel.
REST API + webhook + signed-URL callbacks, 10 GB/month included, per-GB after. EUR468/year option (~EUR39/mo). Token top-ups from EUR5.
On-prem / air-gapped / sovereign-cloud deployment, source license, domain-tuned engine, dedicated engineer. Custom pricing, booked via Cal.com.
OEM bundled, OEM per-platform, source, and strategic license tiers for integrators wanting binary or source rights. See pricing.
Pedulli is founder-financed and bootstrapped. Internal estimates suggest the per-request cost on a low-cost host is small relative to plan pricing, but unit economics are an estimate, not audited profitability.
Bucket B — target market
Per-vertical ratio figures (e.g. telemetry, DICOM, genomics, time-series, AI weights) are internal projections and targets, not delivered customer results.
Bucket C — SRD Engine research tracks (internal, under verification)
The Pedulli SRD Engine is a separate product from the Orchestrator: a deterministic, symbolic, local universal deterministic intelligence framework (PDLI) — built on PDLI, a sealed deterministic symbolic core. It uses no tokens, no context window, no neural net, no machine learning, no LLM — no ChatGPT, Claude, Mistral or Ollama, and no fallback to any model. Each capability is proof-gated: it ships only with a measured benchmark / leaderboard proof object, and research tracks stay labelled research until reproducible. The current tracks — SAT / constraint satisfaction, automated theorem proving, AI / inference acceleration, source-code optimization, and reverse engineering, among others — are listed for transparency, marked under verification, and carry no public win, record, or solved-problem claim. The engine verifies what you supply; it does not break, crack, or solve anything by assertion.
See the research tracks ->Disclosed honestly
Pedulli is open to seed / pre-seed / strategic conversations grounded in measured records and a clear roadmap — not grand theses. Book 30 minutes with Francesco Pedulli, Forli, Italy.