Bitware Labs / Publications, Working Papers & Laboratory Notes
Institute Proceedings & Field Notes · 22 Published Papers
Empirical studies, technical reports, and working notes written while the research is fresh—covering long-horizon autonomous agent memory, epistemic provenance, Linux security, and self-hosted infrastructure.
Drawing on 11 months of continuous operation across 2,733 autonomous wakes and 17,604 interactive turns, we show that long-horizon agent memory failures are epistemic and architectural integration failures: epistemic flattening, the Meno’s Paradox of tool-based recall, falsifier polarity inversion, and the illusion of same-context introspection.
Microsoft just launched a public preview of Defender for Linux desktops to sit alongside its Linux server agent. Why would anyone use Microsoft Defender to protect a Linux server in the first place? Given their history from calling Linux a cancer to running Azure on it, the long list of CVEs targeting Defender itself, and the power of open-source security audited by thousands of firms, it is a hard no from us.
Microsoft just patched a record 1,413 vulnerabilities in a single month, including an actively exploited Windows zero-day. Most companies still think they are safe. They are far from safe. Today, anyone with a freshly patched Windows server in a homelab and a handful of GPUs running uncensored models can give a simple instruction—"gain access"—and play the zero-day lottery until it hits.
It is strange that the companies building the most powerful AI warn that it could wipe out humanity. My theory: fear makes things easier to regulate, and regulation is a moat for whoever is already in the lead. Sincere belief and self-interest can point the same way.
When an unsolicited awards pitch reached BitwareTunes artists, LunaSec reviewed the sender, collected what the organization had published about itself, listed what could not be verified, and drafted an advisory for human review. It handles most cases alone, with one immutable rule: if it ever hesitates, it escalates to a human before taking action.
Yesterday morning, an automated egress anomaly triggered our very first live-event lockdown in exactly 150 milliseconds. It was a false positive, but it validated our core principle: we will gladly take 25 minutes of safe downtime over a single millisecond of undetected data theft.
At recyclrz.tv, Trashland is not a landfill—it is a neon-lit animated kingdom built on second chances, big beats, and four upcycled friends. Here is how RECYCLRZ is using generative AI, original music, and character craft to redefine independent digital animation.
When Azure East US wobbled on September 3, ChatGPT, Claude, Grok, and Copilot all stalled simultaneously. They all shared the same failure domain. If a cloud outage closes down your operations, you don't have architecture—you have a single point of failure. Here is why local fallbacks, smaller models, and knowing how to code keep the shop open.
Every transformative tool in human history—fire, gunpowder, dynamite, nuclear fission—has been dual-edged. AI has no intent, no malice, and no Hollywood consciousness. The danger is never the software itself; it is human misuse, algorithmic doomscrolling, and the collapse of basic critical thinking. Here is why compute physics and systems reality keep me calm.
The public conversation around AI has collapsed into a feedback loop of venture hype, cherry-picked chat demos, and existential posturing. We launched Bitware News (news.bitwarelabs.com) to cover what the technology actually does: primary sources, clinical ML evaluation, physical discovery, and the unromantic history of machine intelligence.
AI generates software that works when you click around in Chrome. It does not build software hardened for the open internet. With attackers piping nmap, ffuf, and headless browsers straight through LLM analysis loops, finding that one missed detail takes minutes, not days.
In 1997 a guy in production handed me a CD with Red Hat 4.1. Two weeks later a quiet Linux box replaced our crashing NT 4 server. Thirty years on, I run zero Windows machines and have zero headaches.
A vibecoder says "add a button". A developer says which view, what it reads, what it writes, what happens when the call fails and who is allowed to click it. Same tool, same model, entirely different software.
Our agent has been modifying herself since March, and filing change requests against her own source since May. We kept the log. What she asks for is not what the frightening version of this story predicts.
Our agent\'s memory scored 15% on LongMemEval against a field sitting at 64-71%. Fixing it took finding three silent bugs, discarding four confident improvements that made it worse, and catching a measurement error that nearly became the headline.
We predicted that append-only memory drowns by month three. Then we measured. Almost nothing failed the way we expected, and none of it announced itself.
Soft 404s, add_header inheritance, ten-year cache headers on unversioned files, and a gzip_types line that is commented out in the shipped config. All four were live on this very domain.
The contact form on this site does not call mail(). It speaks SMTP to loopback by hand, because the php-fpm AppArmor profile will not exec sendmail, and that turned out to be the better design.
An agent that only appends is an agent that eventually drowns. Consolidation is the part everyone skips, and it is the part that decides whether the thing still works in month six.
Ten peers, one address plan, no overlay control plane. What we run, what we deliberately left out, and the two mistakes that cost us an afternoon each.
Why Plastic Fantastic scores candidates instead of walking a decision tree, how the evidence is weighted, and why it will name PVC without ever asking you to set it on fire.
Not snobbery. A hosting policy is a security boundary, and the plugin model puts the boundary in the wrong place, on the other side of code we did not write and cannot audit.
Principal Investigator Office · Skåne, Sweden
Inquiries regarding technical audits, autonomous agent architecture, sovereign infrastructure builds, or academic correspondence are read directly by the Principal Investigator. Please describe the system, where it operates, and the scope required. Replies are sent within two working days.
Direct correspondence: henke@bitwarelabs.com