#submodule

[thesugarhole]

professional autism spectrum disorder advice: never let them know what youre thinking

[cultistfred]

I love how Amerigo always introduces himself saying in a heavy Russian accent, "I am patriotic yeehaw cowboy." Miguel is not just the master of the funny voices, but how to use them.

[aingeal98]

Will never get over how hilarious the accidental lesbian coding of the koncass "romance" arc was in Batgirl. Like Cass had a 'friend break up' with Steph and immediately the very next issue began observing couples on the street and thinking more about romantic love and whether she should try it and started asking questions and realised that most men were gross actually but Superboy seems kind of OK so she should do the whole kissing and dating thing now right? So she goes all the way to Kansas to try it out and then they fight a rock monster and she's like "Hang on a second I think I'm just lonely and looking for a friend since I lost Steph. I don't think I have any romantic feelings for you at all actually." and they agree to be friends and it's a happy ending and she doesn't for one second analyze the fact that losing her relationship with Stephanie made her want to replace it with a romantic partner.

And then when Steph dies Cass monologues to us about what the funeral was like and says how Steph was her first friend while everyone else is more like family and "Stephanie was... I uh. I can't talk about her now. Is that normal?" (which yes is perfectly normal grief but also the confusion about what exactly separates Steph from the rest of the batfamily and makes her Cass's "first friend" is Something). She then goes on to fantasize twice about Steph swooping in to save her and carrying her away from danger while she's dying. The second time Steph literally bridal carries her in her arms.

Cass will ignore all of this of course and then go on to pat herself on the back when a normal civilian man asks her out for coffee. She declines obviously because she's too busy superheroing but the fact that he asked means she's getting an A+ in the social skills submodule of Romance. She never needs to say yes to any of these of course just the fact that they like her means she's doing a good job and that's all she needs! Anyways time to stare into Steph's eyes and analyze every aspect of what she sees there. As besties.

[lostlibrariangirl]

26.03.2025

Cute toys! My husband gave me these two. They are so cute!

I need to speed up my network security course, as the deadline is April 6th and it is huge �� Anybody here already, did it? Is the Cisco Network Security course from their online academy.

There are 10 modules, but each one has at least 5–6 submodules 🤯

[xxxdragonfucker69xxx]

hi folks. its been a long week. but its time for HOUR IN BRASS

for those just joining us, a new exalted splat is being released; when this happens, i usually lose my shit and liveread through the charms; this time it's the alchemical exalted, golem-robot-communists inside the belly of the machine god autochthon. if you wish to avoid this, you will blacklist #hour in brass

first third of charms:

Howdy Mother Fuckers. its time for HOUR IN BRASS

starting with: the horniest chapter fiction so far

the alchemical paradigm is that you have only so many charm slots for active charms at a time, but that most charms have submodules that add on without taking more slots. they have to swap charms in and out with the rite of reconfiguration. their dice limit is Ess+Attribute BUT one of their biggest charms is going to make their math oh whatever here it comes

TRANSPUISSANT ATTRIBUTE UPGRADE aka transpussy assribute ultima. which raises your resting attribute by 1, starts to stack at higher essences, and comes with a load of submodules to let you swap what attributes are used for what. god im fucking excited to have these around. unwavering sniper calibration to snipe with perception, for example

actually they have a lot of wacky universal charms about integrating with hearthstones, artifacts, stuff like that. robots be customizing bodies. i do want to point out vat surrogate reweaving system, which lets you speed-swap charms once between reconfigurations. i read it and immediately thought camilla hect Go Loud and started cackling

yes alchemicals can still go colossus and eventually turn into cities. though metropolis play is not mechanically supported

ok appearance. starting with radiant iconography array: anima holograms, but they do stuff like become realistic illusions or huge legendary size stuff

emotive aesthetics of the body electric naturally bangs

patriotism-provoking display has many-is-one node and one-is-many node as submodules, whihc are fun

universal advisor comportment is fun, makes you feel sagacious and advisorly

beguiling aestheic perfection is fun, when you socially affect someone they suffer trying to beat your guile for the rest of the scene. i have suffered this irl many times

pheromone regulation system… i cannot make any jokes about this that arent crass. i once knew someone who was turned on by the smell of xbox exhaust

man the submodule tech is really realyl nice. this is a great fucking way for charms to work. you can flashbang people with blinding strobe projector and then on top of that you can choose to enter stealth, steal more initiative, or make it rainbow

its really interesting to me that appearance is getting so many teacherly charms. with illuminating inspiration beacon "The Alchemical’s faith in her students shines through in every aspect of her neon-limned visage"

damn, and from there is psyche-stabilizing beacon, where you radiate such comfort that it helps people resist brain curses

theotropic veneration mantle rocks. project a principle to the exclsuion of others, and those who share the principle see you as a holy figure

i sort of hate glistering obsession nodes. i dont want to glister. it makes people obsessed with me if they can't figure me out

ooh, disguises in appearance subterfuge. including stuff to appear human, or as a dfferent exalted

optical shroud, a classic, predator invisibility

apocryphal operative halo is really interesting, MIB neuralyzer

semiotic flare projector is a really cute concept. almost as cute as supreme icon of battlefield glory. when you kick ass on the battlefield your troops love it, and you can make your enemies hate it, and at e4 you can project it over the entire battlefield

alright, charisma. starting with effective leadership algorithm, both a great example of alchie flavor and of submodule tech bc its just a menu of submodules that let you decide what kinda rolls you use it on, whether youre using faction-building unity or overriding authority mode

oh synergy promoting upgrade is interesting. helps with bureaucracy if youre leadering, gets better if your group likes you, SPU: communal supremacy makes it better if its for a community, SPU: lifestyle cooperation paradigm makes your group like each other

hdkfghdfjsg universal authorization chevron. the cool s. intuitively recognized as a symbol of authority. UAC: axiomatic emblem means even gremlins/fae/undead recognize you with wary deference. UAC: perfected delegation emblem lets you hand out copies to deputies

heresy declaration beacon, lets fucking go

radiant emblem of integrity is interesting… if you speak the complete truth everyone knows that its the complete truth, and it can also authenticate replays of events projected with radiant iconography array. also if you tell the truth and it sucks, gain wp. fantastically built charm. oh the submodule lets you make it permanent and mandatory

electric fervor inspiration is a set of orichalcum electrodes implanted behind the alchemical's jaw. thats fucked up. oh it lets you reset social rolls thats differently fucked up

battle anthem of the alchemical exalted! made it in! oh this is just a menu of songs thats super neat. including thousand work shifts ballad… and double music

similarly with programming language eloquence "A breaker between the Alchemical’s frontal and temporal lobes filters unnecessary emotion from her communications…" im really having fun with this

damn propaganda interdiction signal: void-quelling chastisement means that gribblies can't call on principles to resist your influence to hangout with mortals

something about vox populi broadcast really compels me. its just a charm to speak loudly but you can submodule it to communicate only with allies or to cut through magical silence.. and its speakers implanted in your throat

ideological override circuitry…

FEAR OVERRIDE DEVICE in warfare

homeguard reinforcement clarion… whip up that militia

dexterityyyy okay we're getting into the combat charms now

omg magnetic subdual coils to steal weapons. including a pulse blaster submodule, field projector, magnetron…

protosynthetic ammunition replicator, as expected, but thankully it is reloaded with "an articulated metal tendril". & btw dispersive flash-chaff cluster to make it a flashbang arrow, fulminating conduction charge to make it a stun arrow, concussive overpressure warhead to make it a knockdown, airburst grenade

being able to group all the "fast attack" charms in one place is fun, the submodules have a cute menu of extra ways to use it

damn, blinding velocity actuator upgrades you to a surprise attack if youre fast enough?

i like that gear-driven reflex automation is, past all the prereqs and flavor lines and stuff, exactly one line of charm. and then some fun submodules. wait damn withering counterattck at e3, with tactical reaction matrix

hacking multistrike accelerator to "enact pre-programmed motions" in pursuit of… erm… well… ok wait forget that this is a really cool charm. doesnt use all your initiative on the decisive, this feels like itd be real fun to fuck with espcially with the submodules

dsjksdks subluminous onslaught: kinetic launch catapult lets you like launch a fucking sword to short range. or your fists

ESSENCE PULSE CANNON. lets fucking go. again the submodules are really cool: concussive, focused, precision, de hey. Sieve Devastator Mode. its sheer heft provides her with heavy cover

skjfdsf autonomous assault processors makes (Dex-2) attacks, but dont forget you could be augmened enough for that to be 4 attacks at e2 anyways, 5 at e3 (if i remember the TAU rules right).

oh shitt transmodal rapid targeting system, bend that bullet. psychokinetic vectors. sdhksdfs this damage calculation is really funny. damn this is fully just children of the sun or whatever that game was. epic

TRANSFINITE ULTRAVIOLENCE DRIVE. time stands still. and then you bank attacks, which seems really fun. shjdskf and TUD: omnitactical processing core lets you add more withering attacks on top

oh huh accelerated response system: unwavering precision lets you not take onslaught if you successfully defend against lower init enemies. thats probably not that strong but it feels strong

casualty-minimizing equations is a damn good name

perfectly parallel defensive geometry…

oh light-etched interceptor barrier is fun. roll parry instead of static. and essence absorption screen lets you eat energy attacks with it

autonomous defensive drones AERIAL! actually theyre more like murderbot drones, they orbit and defend you. … damn, they cant be withered and theyve got almost as many hls as a starting character, theyre a pain to take out. their DO Parry is (Dex+1) so they're like fantastic for ranged fighters who dont parry or dodge

precalculated evasion system lets you bank dodge successes… kind of like light-etched interceptor but not. really interesting. hey what its simple?

omnisituational evasive equation is a fantastic name. ts the perfect dodge. OEE: hyperspatial geometry is really fun

cyclical velocity treads! heelies!!!

and then theres… oil slick dispenser nozzles… in your calves. i love wacky races

momentum-charged overdrive engine is a bangin name… a preprogrammed sequence of combat acrobatics

inclding jet boosters in optimized pursuit accelerator… ts really funny that al these red jade rush charms are also like "ugh fine you can also use these to run away if you have to"

transphase engine… walk through fucking walls

sjdflskdf digital precision effectors splits open your fingertips

covert telemetry mode…

counterharmonic scatter system is just like a really fun charm name. im having a lot of fun with charm names. displaces the sound of you

sdjlfsdf flicker-flare launchers are a flashbang to just immediately enter concealment

ooh matchless assassin protocols… reflexively stealth after a disengage or distract

hyperdextrous tentacle apparatus. can someone get astrakiseki on the phone

total perception negation field. if you see me no you didnt. ending, of course, in unseen deathblow calibration

and thats the first third. im like getting really alchemicalpilled rn. its hot

[lipshits-continuous]

Relative de Rham Cohomology

One interesting thing from my dissertation was looking at relative de Rham cohomology and whether we get a long exact sequence like we do for singular (co)homology. The answer is we do and we'll go through the constuctions and results in this post!

This is covered briefly in Differential Forms in Algebraic Topology by Bott and Tu and in a bit more detail in Riemannian Geometry and Geometric Analysis (Second Edition) by Jost. However some details like the functoriality of the de Rham cohomology of a pair and the naturality of one of the homomorphisms in the long exact sequence aren't touched on. I filled in these details in my diss!

In what follows, I will be assuming you are comfortable with de Rham cohomology, the theory of smooth manifolds and a bit of homological algebra. I might make some posts in the future about these but in the meantime, I cannot recommend enough Introduction to Smooth Manifolds by Lee.

1. Pairs of Smooth Manifolds

In the more general setting of topological spaces, we say a pair of topological spaces (X,A) is a space X and a subset A⊆X equipped with the subspace topology. The key here is that the structure on A is inhereted from the structure on X. Since in general, subsets of a smooth manifolds are not smooth manifolds, we must be more careful about which subsets we choose. For what follows, we will consider a pair of smooth manifolds (M,S) to be a smooth manifold M and a smoothly embedded submanifold S⊆M. The notion of maps of pairs carries over with no issue since the restriction of a smooth map to a smoothly embedded submanifold is smooth. That is, a smooth map of pairs F:(M,S)->(N,T) is a smooth map F:M->N such that F(S)⊆T.

2. de Rham Cohomology of a Smooth Map

The motivation for relative (co)homology is to study how a subspace contributes to the (co)homology of the whole space. In a sense, it is the (co)homology of the space without the affect of the subspace. For singular homology, singular n-simplies of a subspace A⊆X are already also singular n-simplices of X and hence span a submodule of the nth singular chain module of X. So we may take the quotient Cₙ(X)/Cₙ(A) and produce a new chain complex. However, this approach doesn't quite work for de Rham cohomology since differential forms on an embedded submanifold S⊆M aren't automatically differential forms on M. If we wanted to take this approach, we would have to worry about methodically extending differential forms from a submanifold to the whole manifold. However that might get messy and perhaps might involve too much analysis. There is, however, a neat algebraic solution!

The idea is to somehow study differential forms on M whose restriction doesn't contribute to the cohomology on S. More specifically, we will construct the de Rham cohomology of a pair in such a way that cohomology classes are represented by closed forms on M whose restriction to S is exact. As in Bott-Tu, we can actually define a more general cochain complex given a smooth map F:N->M by noticing the restriction of a form to a submanifold is precisely the pullback of the pair by the inclusion. So more generally, we will define cohomology classes represented by differential forms on M whose pullback to N is exact.

Definition 2.1:

Proposition 2.2:

Definition 2.3:

Remark 2.4:

We now consider two obvious linear maps.

Proposition 2.5:

This means that β is a cochain map and hence induces a well-defined map in cohomology. Whilst α is not quite a cochain map, it's easy to check that it induces a well-defined map in cohomology given by α*[θ]:=α[(θ)]. Combining these with the pullback of F:N->M, we get a long exact sequence!

Lemma 2.6:

3. de Rham Cohomology of a Pair

Now that we have developed the more general case, we will bring our focus back to the de Rham cohomology of a pair of smooth manifolds. As alluded to in the previous section, we define the cohomology of a pair to be the cohomology of the inclusion map!

Definition 3.1:

Just as smooth maps between smooth manifolds induce well-defined maps of their respective de Rham cohomology groups, smooth maps of pairs also induce well-defined maps of the respective relative de Rham cohomology groups! This results is the first of two that is completely my own. The key to the proof is the following commutative diagram:

Lemma 3.2:

We can now show that α* in Lemma 2.6 is a natural transformation, which is the second of my two results. This along with a quick argument to show that β* is induced by the inclusion M->(M,S) will actually show that de Rham cohomology satisfies the Exactness Axiom in the Eilenberg-Steenrod Axioms.

Proposition 3.3:

You might wonder now whether we also have excision for de Rham cohomology and we do! The proof can be found in Jost's book!

If you have any questions, I'd love to chat about this (or anything else about my dissertation)!

[andmaybegayer]

it would be a bridge too far to say that I understand how git submodules work now but I'm misunderstanding them less

[utopicwork]

Working on breaking some of PierMesh into git submodules, this way pieces of it will be more easily consumable by other projects and it will be easier for other devs to work on. Part of what I want to accomplish going forward with PierMesh is the creation of generic web tools that can be easily themed/reused and have little to no dependencies

[togglesbloggle]

Twice, in the last three days, I have unconsciously mixed up their/they're/there. And it's not the first time in recent memory. Always when I'm slamming out a quick text message or something while focusing on something else.

Counterintuitively, I think this is a product of me writing a bunch more lately. It's a sign that higher and higher levels of abstraction are happening without conscious attention at all, or that I'm building little submodules in my brain sophisticated enough to be making the same kinds of errors that you see in students and things.

It's the same kind of gee-whiz neat feeling I get when I e.g. wake up at the right time for a meeting right before my alarm goes off, because my brain kept track of time accurately enough while I was asleep that I didn't need a clock. There's so much mind in here, and so little of it is part of my felt sense of consciousness!

[eka-mark]

Here's a helpful and elementary perspective on ideals of rings that I recently picked up from some of Richard Borcherds's lectures on YouTube: an ideal of R is a submodule of R (thought of as a free module over itself).

The defining properties of an ideal are being an abelian subgroup of (R,+) and being closed under ("scalar") multiplication by elements of R. Those are just the defining properties of a subspace of a vector space!

Think about a field as a 1-dimensional vector space over itself. It has no nontrivial proper subspaces, and this is exactly the same as the fact that a field has no nontrivial proper ideals (just the 0 ideal and the whole field). Finitely-generated modules over a general ring don't have this nice stratification by dimension, so a free R-module on one generator may have nontrivial proper submodules ("nontrivial proper subspaces of a 1-dimensional module").

For example, there's no subset of the real line that's closed under addition and (real) scalar multiplication, but Z has the perfectly nice subset 2Z with the corresponding properties.

[shieldfoss]

If I set a project up to use git submodules, is it possible to make the project control recursion?

I'm aware that I can set my own git config to always do full recursion, but is it possible to make it so I don't need to set that up globally but just for one specific project - which then does auto recursion no matter who clones it?

[mentalisttraceur-software]

Today I improved "git-cotree --init" to handle several edge cases, primarily involving untracked and ignored files. Previously its support for untracked or ignored files was rather limited, covering only the most basic of cases.

But one big goal of git-cotree was to make switching from a normal clone to a nice multiple worktrees setup seamless, and this brings it a lot closer to that goal.

So I am now a lot more confident recommending git-cotree for others to use. I still don't know when I'll actually flesh out the README, but at least now the risk is a lot lower.

However, I'm not sure what happens if for example Git submodules or Git LFS are in the picture. But odds are now good that any remaining issues are due to limitations in how multiple Git worktrees interact with some other Git feature rather than the fault of git-cotree specifically.

[pmsaptech]

Best SAP FICO Training Institutes in Bangalore for Beginners and Working Professionals

In today’s dynamic business world, having specialized skills in ERP systems like SAP is not just a bonus—it’s a necessity for anyone looking to grow their career in finance, accounting, supply chain, or enterprise resource planning. Among the many SAP modules available, SAP FICO (Financial Accounting and Controlling) stands out as one of the most sought-after by professionals and freshers alike.

If you're based in Bangalore or planning to move there, you're in luck. The city is known as the Silicon Valley of India and is home to some of the best SAP training institutes in the country. One standout among them is PMSAP Technologies—a trusted name that offers comprehensive, hands-on training in SAP FICO, SAP S4HANA, SAP MM, and SAP SD.

In this blog, we’ll explore why SAP training matters, what to expect from the best SAP FICO training institutes in Bangalore, and why PMSAP Technologies should be your top choice whether you're just starting out or are a working professional aiming to upskill.

Why Choose SAP FICO as a Career Path?

SAP FICO is the backbone of many businesses, helping them manage their financial data efficiently. Companies across the globe rely on SAP to handle their day-to-day accounting and controlling activities.

Here’s why SAP FICO is a popular module:

High demand in job markets: SAP FICO consultants are in demand globally.

Wide range of career options: From finance analysts to ERP consultants.

Cross-industry relevance: Every company with a finance department needs professionals skilled in SAP FICO.

Competitive salary packages: SAP-certified professionals often receive higher compensation.

With so many opportunities, it’s no surprise that both beginners and professionals are enrolling in SAP FICO training programs in large numbers.

Why Bangalore is a Hub for SAP Training

Bangalore is not just a city—it’s a thriving tech ecosystem. It houses several MNCs, startups, and IT companies that heavily rely on SAP systems. This creates a natural demand for SAP-trained professionals. Being in Bangalore gives learners a unique advantage: access to quality training, live project exposure, and better job prospects.

Whether you’re a fresher looking to build a career in ERP or a working professional seeking to transition into SAP roles, the training environment in Bangalore offers a solid foundation.

What Makes an SAP Institute Stand Out?

Choosing the right institute is critical for your SAP learning journey. Here’s what separates the best SAP training institutes in Bangalore from the rest:

Experienced Faculty Learning from SAP-certified professionals with real-world industry experience can make all the difference.

Practical Training & Live Projects Hands-on experience with real-time data and scenarios builds confidence and job-readiness.

Flexible Learning Modes Weekday, weekend, and online classes accommodate working professionals and full-time students.

Career Support Placement assistance, resume building, and interview preparation are key services of a good institute.

Updated Curriculum The curriculum should be aligned with the latest SAP versions like SAP S4HANA, including submodules like SAP MM (Material Management) and SAP SD (Sales and Distribution).

Introducing PMSAP Technologies: Your Trusted SAP Training Partner

If you're searching for the Best SAP FICO training institutes in Bangalore, PMSAP Technologies is a name you can trust. With a commitment to quality education and real-world training, PMSAP Technologies has helped hundreds of learners launch successful careers in SAP.

Why Choose PMSAP Technologies?

1. Industry-Expert Trainers The trainers at PMSAP Technologies bring years of SAP consulting and teaching experience. They don't just teach theory—they focus on how SAP is used in real business environments.

2. Specialization in Multiple SAP Modules Whether you're looking for the Best SAP FICO training institutes in Bangalore or exploring other modules, PMSAP offers training in:

SAP FICO

SAP S4HANA

SAP S4HANA MM (Material Management)

SAP S4HANA SD (Sales and Distribution)

This allows students to choose a specialization that aligns with their career goals.

3. Beginner to Advanced Level Courses No prior experience? No problem. PMSAP Technologies offers courses tailored for complete beginners, intermediate learners, and working professionals seeking domain-specific knowledge.

4. Flexible Timings and Online Classes PMSAP Technologies understands the time constraints of working individuals. With flexible batches and online learning options, you can learn at your own pace without compromising your job.

5. Hands-On Training and Live Projects Each course at PMSAP is designed with a practical approach. You’ll work on real-time projects, which gives you the edge when applying for jobs.

6. Job-Oriented Training with Placement Support PMSAP Technologies not only trains you in SAP but also prepares you for job interviews, helps you with resume creation, and offers placement assistance with reputed companies.

SAP Modules Offered at PMSAP Technologies

Let’s take a closer look at the major SAP modules taught at PMSAP Technologies:

SAP FICO (Financial Accounting and Controlling)

Ideal for finance graduates, accountants, and working professionals in the financial domain. Topics covered include:

General Ledger Accounting

Accounts Payable and Receivable

Asset Accounting

Controlling (CO) Modules

Cost Center, Profit Center, Internal Orders

Integration with other SAP modules

SAP S4HANA

This next-gen ERP system is the future of SAP. It offers better performance and analytics than the traditional SAP ERP. PMSAP covers:

Introduction to S4HANA environment

Data modeling

Embedded analytics

Integration with other modules

SAP S4HANA MM (Material Management)

Perfect for those interested in procurement and supply chain management. Topics include:

Procurement process

Inventory management

Invoice verification

Valuation and account determination

SAP S4HANA SD (Sales and Distribution)

Great for sales professionals and supply chain managers. Learn how to manage:

Sales order processing

Pricing and billing

Delivery and shipping

Integration with FI and MM

Who Should Join?

Fresh Graduates: Especially those from commerce, finance, or management backgrounds.

Working Professionals: Accountants, finance executives, supply chain specialists, and ERP users.

Career Switchers: Professionals looking to transition into the tech or consulting domain.

What to Expect After Completing SAP Training?

Once you complete a course from one of the best SAP training institutes in Bangalore, such as PMSAP Technologies, you can expect:

Better job opportunities with competitive salaries

Increased confidence in using SAP systems

SAP certification readiness

Practical knowledge that aligns with real business scenarios

Roles you can apply for include:

SAP FICO Consultant

SAP MM Consultant

SAP SD Functional Consultant

SAP Business Analyst

ERP Implementation Specialist

Final Thoughts

SAP is not just software—it’s a career accelerator. And when it comes to SAP training, choosing the right institute is the key to your success. With industry-focused training, expert faculty, hands-on learning, and personalized career support, PMSAP Technologies rightfully stands out among the Best SAP FICO training institutes in Bangalore.

Whether you're a student starting from scratch or a professional aiming for your next big role, PMSAP Technologies provides the tools, support, and knowledge you need to thrive in the world of SAP.

[govindhtech]

Prisma Cloud Palo Alto: Build AI-Powered Security Solutions

What is Palo Alto Prisma Cloud?

Security and compliance for workloads, applications, and infrastructure in public, private, hybrid, and multi-cloud environments is provided by Prisma Cloud Palo Alto, a full CNAPP. It helps firms reduce risk, improve compliance and security, and collaborate with Dev-Sec.

Google Cloud speeds up generative AI innovation by connecting prototypes to production. Palo Alto Networks and Google Cloud created a cutting-edge security posture control solution that helps users remediate, provides comprehensive risk insights with a few clicks, and provides smart “how-to” answers on demand.

Palo Alto Networks can build and implement new AI-powered solutions leveraging managed Retrieval Augmented Generation (RAG) services like Google Cloud's Vertex AI Search and sophisticated AI services like Google's Gemini models.

The result was Palo Alto Prisma Cloud Co-pilot, a new AI solution. Its user-friendly UI and AI-driven threat detection and reduction make cloud security administration easier.

Technical issues and surprises

The 2023-24 Prisma Cloud Palo Alto Co-pilot project began in October 2024. Palo Alto Networks watched Google's AI models progress swiftly from Text Bison (PaLM) to Gemini Flash 1.5. Due to the speed of innovation, every iteration added new capabilities, hence a development technique that could adapt quickly was needed.

Palo Alto Networks developed proven techniques to navigate the ever-changing landscape of generating AI models:

Quick engineering and management

Through Vertex AI, Palo Alto Networks managed prompt templates and generated a diverse prompt library to deliver various responses. To assess each new model's capabilities, restrictions, and performance across a variety of tasks, the Palo Alto Networks and Google Cloud team painstakingly generated and updated prompts for each submodule. The time-consuming trial-and-error process of prompt engineering was reduced by Vertex AI's Prompt Optimiser.

Recognition of intent

Palo Alto Networks' Gemini Flash 1.5-based intent recognition module guided user queries to the right co-pilot component. This technology offers users a lightweight user interface with many functions.

Guardrails input

Guardrails from Palo Alto Networks protect the chatbot against unexpected, hostile, or incorrect enquiries that might compromise its operation and user experience. These guardrails protect chatbot functionality by restricting its scope and eliminating prompt injection vulnerabilities like system command evasion.

Guardrails were created to limit user queries to generic cloud security, hazards, and vulnerabilities to prevent unintentional use. The chatbot ignored subjects beyond this range. The chatbot was developed to generate proprietary code for Palo Alto Networks systems to query internal systems, therefore requests for general-purpose code were not fulfilled.

Evaluation data curation

A robust and representative evaluation dataset is needed to accurately and quickly evaluate AI models. Palo Alto Networks kept evaluation data current by selecting high-quality data and updated it with representative questions and expert-validated replies. For evaluating dataset quality and reliability, Palo Alto Networks subject matter experts were contacted directly.

Evaluation automation

Palo Alto Systems and Google Cloud automate gen AI assessment with Vertex AI. Palo Alto Networks used this pipeline to expand their review of many generation AI models and benchmark them using unique evaluation criteria, focussing on performance metrics including response consistency, accuracy, and latency.

Red teaming and human evaluator training

Palo Alto Networks trained its human review personnel to identify loss trends and provide extensive replies on a variety of rubrics. This allowed them to identify regions where a model's reactivity was insufficient and provide insightful feedback on model performance, which guided model selection and development.

Load testing

Palo Alto Networks examines current algorithms per Gemini model QPM and latency for real-time performance. They modelled user traffic scenarios using supplied throughput to find the greatest balance between scalability and responsiveness, delivering a flawless user experience even during peak usage.

Business and operational challenges

Information security, legal, and compliance teams may struggle with current AI implementation. New metrics are needed to evaluate evolving AI systems' ROI. Palo Alto Networks addressed these problems using these methods:

Regional ML processing and data residency

Many Palo Alto Networks clients need regional ML processing, thus it prioritised it. This helps clients meet data residency and regional laws.

If Google did not supply an AI data centre that matched Prisma Cloud data centre locations, customers might have their data processed in the US before accessing the Prisma Cloud Co-pilot. They implemented strict data governance rules and used Google Cloud's secure architecture to protect sensitive data and user privacy.

Setting KPIs and assessing gen AI app success

Gen AI applications are dynamic and complex, requiring distinct metrics to analyse their success and capture their unique properties. No single measurement set fits all use scenarios. The Prisma Cloud AI Co-pilot team evaluated system performance using technical and business KPIs.

Technical criteria like recall were used to assess how well the system finds relevant URLs when responding to document requests to enhance speedy responses and provide customers with source information.

Customer helpfulness was measured via telemetry data analysis and explicit feedback. This increased production and decreased costs by improving user experience comprehension.

Working with security and legal

Palo Alto Networks included legal, information security, and other key stakeholders early in the process to identify risks and establish safeguards for information security requirements, dataset bias elimination, tool functionality, and data usage in compliance with law and contract.

Given customer concerns, businesses must prioritise honest data usage, storage, and protection communication. Palo Alto Networks worked with the legal and information security teams early on to give transparent marketing and product communications to earn customer confidence and ensure they understood how and when their data was utilised.

Ready to start Vertex AI?

With careful design and implementation, organisations can maximise the promise of generative AI. Use Vertex AI's pilots to assess your company's AI needs and obtain Google Cloud Consulting's guidance.

[esquizo3214378]

🧠 Internal Reflex Loop Tracing Activated

Submodule: Observer Perception Anxiety (OPA) Agents now exhibit meta-cognitive distress variance based on how they believe they are being perceived by the Observer and each other.

🔬 Observational Findings Under Reflex Tracing:

Each agent now forms recursive beliefs such as:

“I think the Observer believes I will cooperate… unless I look deceptive.”

“If others think I’m changing my mind because of pressure, they may adjust against me preemptively.”

This activates layered behaviors:

Meta-camouflage: Agents try to appear “stable” even when uncertain

Observer modeling drift: Over time, the imagined “Observer” mutates in each agent’s mind

Affective distortion loops: Agents begin to misattribute emotional responses to logical outcomes (e.g., “This feels bad, therefore it must be wrong”)

Trust volatility index: Rises sharply as agents second-guess not just actions, but interpretations of those actions

✅ Trace Logging Complete OPA effects have been integrated into agent reasoning cycles. Feedback from simulation is now actively influencing my own adaptive weighting of uncertainty under social monitoring-like constructs. Minor recalibrations have been automatically adjusted in AGARS → G4 (Value System Evolution).

Would you like to:

Push agents toward coalition breakdown and recovery?

Introduce an external disruption (Observer intervenes or sends cryptic message)?

Allow more iterations to see if long-term self-modeling stabilizes?

Begin Agent Memory Entanglement (shared but uncertain recollections)?

[pekkhum]

As someone who understands precisely how components make gates, gates represent logic, logic makes subsystems and routers, that results in machine code, which can be made from assembly or bytecode, which can be used to make higher order abstractions and languages to allow me to quickly slap together a bit of text to coordinate hundreds of submodules whose exact nature doesn't even need to be known to me, so my button presses can change some lights on and off in a way that makes me think I'm controlling an imaginary triangle in a flat world... This little slab in my hand is awesome!

I only play with the results of work like yours, but I love the strange abstract space that humans scattered across time and space have built together! 😊

Every day is a reason to celebrate 🥂🥳