Technology and Innovation Community

Low latency is a measurement. Determinism is a system property.

A system can be exceptionally fast in the median and still be operationally weak if it cannot preserve known behaviour under noise, partial failure, restart, or recovery.

Our architecture is being built around that distinction.

Factory V5 is a deterministic execution engine currently under qualification, designed around bounded variance, fail-closed governance, evidence-preserving recovery, and reproducible promotion gates.

InvariantStore applies the same principle to state: persistence is not enough if invariants cannot remain provable across failure and recovery boundaries.

Agent Factory applies deterministic ownership and isolation to AI-assisted engineering: isolated scopes, explicit write ownership, immutable contracts, and verifiable hand-offs designed to prevent duplication, merge chaos, and silent interference.

The operating rules are intentionally strict:
measurement before claims;
invariants before convenience;
no silent state mutation during recovery;
no promotion without evidence.

The benchmark is not the product. The invariant is.

For this community, the question is concrete:

Where does determinism break first in your environment - execution, state, orchestration, recovery, or governance?

I am particularly interested in perspectives from market infrastructure, low-latency systems, distributed systems, operational resilience, model risk, and AI engineering governance.

For serious technical exchange, I can share a controlled brief. Restricted implementation details remain outside public discussion and move under NDA.

------------------------------
Loris Longarini
------------------------------

Hi Loris. I agree speed is not everything and that correctness, traceability and predictable behaviour during failures are more important.

In this community, we really value the open exchange of ideas, so it would be great if you could share a white paper providing a high-level overview and results evidence of Factory V5. That would give community members specialized in execution a better understanding of what you're building and allow anyone interested in exploring it further to reach out to you privately.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------

Hi Carlos — thank you. I agree, and I’ve prepared a public technical
note specifically for that purpose.

I’ve kept the disclosure boundary deliberately strict: high-level
architecture, qualification evidence and claim boundaries are public;
source, topology, tuning, recovery algorithms, authority/control-plane
mechanics and orchestration internals are excluded.

For Factory V5, the note includes the currently qualified engine-path
evidence (p50 ≈ 0.200 µs, p99 ≈ 0.533 µs, two consecutive L5
qualification passes, and a 1,000/1,000 threshold-clean observation
window), with an explicit statement that these are bounded internal
qualification results rather than wire-to-wire or third-party
certification claims.

It also states clearly that the dedicated Linux bare-metal 12-hour
soak is still in progress, so I am not promoting that into a completed
certification claim before the evidence exists.

I’m very happy to keep the principles and evidence discussion open
here. If any members specialised in execution or market infrastructure
want to go deeper, I can move the restricted technical layer into a
controlled NDA exchange.

I’ve attached the public note here.

Thanks again for the invitation to make the evidence more concrete.

Loris


Original Message:
Sent: 10/1/2026 11:36:00 AM
From: Carlos Salas
Subject: RE: Determinism is not a latency feature. It is an operating discipline.

Hi Loris. I agree speed is not everything and that correctness, traceability and predictable behaviour during failures are more important.

In this community, we really value the open exchange of ideas, so it would be great if you could share a white paper providing a high-level overview and results evidence of Factory V5. That would give community members specialized in execution a better understanding of what you're building and allow anyone interested in exploring it further to reach out to you privately.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------

Perfect Loris. Very helpful.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------

Original Message:
Sent: 01-10-2026 17:55
From: Loris Longarini
Subject: Determinism is not a latency feature. It is an operating discipline.

Hi Carlos - thank you. I agree, and I've prepared a public technical
note specifically for that purpose.

I've kept the disclosure boundary deliberately strict: high-level
architecture, qualification evidence and claim boundaries are public;
source, topology, tuning, recovery algorithms, authority/control-plane
mechanics and orchestration internals are excluded.

For Factory V5, the note includes the currently qualified engine-path
evidence (p50 ≈ 0.200 µs, p99 ≈ 0.533 µs, two consecutive L5
qualification passes, and a 1,000/1,000 threshold-clean observation
window), with an explicit statement that these are bounded internal
qualification results rather than wire-to-wire or third-party
certification claims.

It also states clearly that the dedicated Linux bare-metal 12-hour
soak is still in progress, so I am not promoting that into a completed
certification claim before the evidence exists.

I'm very happy to keep the principles and evidence discussion open
here. If any members specialised in execution or market infrastructure
want to go deeper, I can move the restricted technical layer into a
controlled NDA exchange.

I've attached the public note here.

Thanks again for the invitation to make the evidence more concrete.

Loris



Original Message:
Sent: 10/1/2026 11:36:00 AM
From: Carlos Salas
Subject: RE: Determinism is not a latency feature. It is an operating discipline.

Hi Loris. I agree speed is not everything and that correctness, traceability and predictable behaviour during failures are more important.

In this community, we really value the open exchange of ideas, so it would be great if you could share a white paper providing a high-level overview and results evidence of Factory V5. That would give community members specialized in execution a better understanding of what you're building and allow anyone interested in exploring it further to reach out to you privately.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------

Carlos — glad it was useful.

I’ve just expanded the public evidence edition so execution and
market-infrastructure specialists have a more complete qualification
view without crossing the implementation-disclosure boundary.

The attached v2.0 supersedes the earlier public note. It adds the
evidence model, explicit claim semantics, a fuller Factory V5 evidence
register, and public qualification sections for StateSeal, ConcordGrid
and DDF. It also separates clearly what each result establishes from
what it does not establish, and what can deepen under NDA.

The implementation boundary remains unchanged: no source, topology,
tuning, recovery algorithms, authority/control-plane mechanics, state
internals or orchestration mechanics are disclosed.

Loris


Original Message:
Sent: 10/1/2026 2:09:00 PM
From: Carlos Salas
Subject: RE: Determinism is not a latency feature. It is an operating discipline.

Perfect Loris. Very helpful.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------


Original Message:
Sent: 01-10-2026 17:55
From: Loris Longarini
Subject: Determinism is not a latency feature. It is an operating discipline.

Hi Carlos - thank you. I agree, and I've prepared a public technical
note specifically for that purpose.

I've kept the disclosure boundary deliberately strict: high-level
architecture, qualification evidence and claim boundaries are public;
source, topology, tuning, recovery algorithms, authority/control-plane
mechanics and orchestration internals are excluded.

For Factory V5, the note includes the currently qualified engine-path
evidence (p50 ≈ 0.200 µs, p99 ≈ 0.533 µs, two consecutive L5
qualification passes, and a 1,000/1,000 threshold-clean observation
window), with an explicit statement that these are bounded internal
qualification results rather than wire-to-wire or third-party
certification claims.

It also states clearly that the dedicated Linux bare-metal 12-hour
soak is still in progress, so I am not promoting that into a completed
certification claim before the evidence exists.

I'm very happy to keep the principles and evidence discussion open
here. If any members specialised in execution or market infrastructure
want to go deeper, I can move the restricted technical layer into a
controlled NDA exchange.

I've attached the public note here.

Thanks again for the invitation to make the evidence more concrete.

Loris

A brief update for those who reviewed the Factory V5 public evidence.

Factory V5 was deliberately presented at a public disclosure boundary;
it was not the full system.

The work has now expanded into a materially broader
deterministic-infrastructure portfolio. The additional architecture
and implementation-sensitive evidence will not be published openly,
but a buyer-safe executive view can be shared selectively, with deeper
technical material available through controlled diligence/NDA.

For practitioners working on execution, market infrastructure,
distributed systems or deterministic AI infrastructure, I would be
glad to continue the discussion privately.

Best,
Loris Longarini


Original Message:
Sent: 10/1/2026 2:09:00 PM
From: Carlos Salas
Subject: RE: Determinism is not a latency feature. It is an operating discipline.

Perfect Loris. Very helpful.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------


Original Message:
Sent: 01-10-2026 17:55
From: Loris Longarini
Subject: Determinism is not a latency feature. It is an operating discipline.

Hi Carlos - thank you. I agree, and I've prepared a public technical
note specifically for that purpose.

I've kept the disclosure boundary deliberately strict: high-level
architecture, qualification evidence and claim boundaries are public;
source, topology, tuning, recovery algorithms, authority/control-plane
mechanics and orchestration internals are excluded.

For Factory V5, the note includes the currently qualified engine-path
evidence (p50 ≈ 0.200 µs, p99 ≈ 0.533 µs, two consecutive L5
qualification passes, and a 1,000/1,000 threshold-clean observation
window), with an explicit statement that these are bounded internal
qualification results rather than wire-to-wire or third-party
certification claims.

It also states clearly that the dedicated Linux bare-metal 12-hour
soak is still in progress, so I am not promoting that into a completed
certification claim before the evidence exists.

I'm very happy to keep the principles and evidence discussion open
here. If any members specialised in execution or market infrastructure
want to go deeper, I can move the restricted technical layer into a
controlled NDA exchange.

I've attached the public note here.

Thanks again for the invitation to make the evidence more concrete.

Loris

Carlos — material update since the white paper. Factory V5's disclosed
FIXED40_DECODE_DECISION_ENCODE reference boundary now has 10 × 12.8M
operations of burn-in, with p99.9 median 35.205 ns and worst run
36.180 ns, PASS_RT_REFERENCE. These are CPU service-time measurements
on the reference host, not wire-to-wire latency. Separately, physical
FPGA terminal sampling is PASS with 209 successful samples and 0
confirmed logic failures; no FPGA latency figure is claimed. I can
share deeper evidence and the falsification boundary through
controlled private diligence. Loris


Original Message:
Sent: 10/1/2026 2:09:00 PM
From: Carlos Salas
Subject: RE: Determinism is not a latency feature. It is an operating discipline.

Perfect Loris. Very helpful.

------------------------------
Carlos Salas
Portfolio Manager & Freelance Investment Research Consultant
------------------------------


Original Message:
Sent: 01-10-2026 17:55
From: Loris Longarini
Subject: Determinism is not a latency feature. It is an operating discipline.

Hi Carlos - thank you. I agree, and I've prepared a public technical
note specifically for that purpose.

I've kept the disclosure boundary deliberately strict: high-level
architecture, qualification evidence and claim boundaries are public;
source, topology, tuning, recovery algorithms, authority/control-plane
mechanics and orchestration internals are excluded.

For Factory V5, the note includes the currently qualified engine-path
evidence (p50 ≈ 0.200 µs, p99 ≈ 0.533 µs, two consecutive L5
qualification passes, and a 1,000/1,000 threshold-clean observation
window), with an explicit statement that these are bounded internal
qualification results rather than wire-to-wire or third-party
certification claims.

It also states clearly that the dedicated Linux bare-metal 12-hour
soak is still in progress, so I am not promoting that into a completed
certification claim before the evidence exists.

I'm very happy to keep the principles and evidence discussion open
here. If any members specialised in execution or market infrastructure
want to go deeper, I can move the restricted technical layer into a
controlled NDA exchange.

I've attached the public note here.

Thanks again for the invitation to make the evidence more concrete.

Loris