FlashFeed

Researchers have demonstrated that a computer worm powered by artificial intelligence (AI) can autonomously spread across a network by identifying and exploiting vulnerabilities on different devices, raising fresh concerns about how the technology could change the future of cyberattacks.The proof-of-concept malware, developed by researchers at the University of Toronto and cybersecurity firm CleverHans, combines a locally running large language model (LLM) with an autonomous software agent that can scan networks, assess potential attack paths, and decide how to compromise new targets without human intervention. The researchers say the work shows how AI could enable malware to adapt to unfamiliar environments rather than relying on a single preprogrammed exploit.In experiments described in a new study uploaded June 2 to the arXiv preprint server, the worm was tested against a simulated corporate network containing 33 hosts, including Linux servers, Windows workstation computers and other internet-connected (IoT) devices. The researchers found that the system identified vulnerabilities, compromised new machines, and replicated itself across roughly 62% of the network over the course of a week."The main finding is that this type of system can do more than run a fixed exploit; it can examine the target environment, reason about possible vulnerabilities, use tools to attempt attacks, and then replicate itself after a successful compromise," Michael Agee, an adjunct professor of information technology at Trinity Washington University in Washington, D.C., who was not involved in the research, told Live Science.How does the AI worm work?The setup was relatively straightforward. The researchers took an open-weight LLM (for which training data is publicly available) running on local hardware and connected it to a software framework that could scan networks, collect information about target systems, and carry out attacks. The AI's role was to interpret what it found and decide where to go next."The AI-driven part of the attack is mainly the reasoning and decision-making," Agee said. "The LLM is not magically hacking the system; it is being used to reason about what the information means, suggest possible attack strategies, decide which tool or action should be tried next, and help adjust the approach when something fails."Intelligence does not exist in discovering new vulnerabilities; rather, intelligence exists in determining how quickly an attacker can choose and sequence attacks against previously identified vulnerabilities.Bob Hutchins, adjunct faculty at Lipscomb UniversityIn other words, the worm isn't inventing new ways to break into systems. Instead, it's taking information about a machine, matching it against known vulnerabilities and weaknesses, and deciding which avenue is most likely to succeed.Bob Hutchins, who teaches AI strategy courses at Lipscomb University in Nashville, Tennessee, said the innovation lies in the system's ability to adapt."Traditional worms follow a scripted sequence: Once a vulnerability is identified, the worm replicates," Hutchins told Live Science. "In contrast, the researchers demonstrated that an easily downloaded AI model could be used as the decision-making component of the worm. The worm would analyze each device it encountered to determine its most effective strategy to breach that particular system.""Intelligence does not exist in discovering new vulnerabilities; rather, intelligence exists in determining how quickly an attacker can choose and sequence attacks against previously identified vulnerabilities," he added.What makes this AI worm different from conventional malware?The researchers also designed the worm to work across devices with different levels of computing power. More capable compromised machines equipped with graphics processing units (GPUs) could provide reasoning services for lightweight agents running on less-powerful devices elsewhere on the network."What made it particularly dangerous was a clever tiered design," Tom Vazdar, a professor of AI and cybersecurity at the Open Institute of Technology, told Live Science. "GPU-equipped compromised machines provided reasoning capacity for lightweight agents running on low-power IoT devices that couldn't run an AI model locally. A camera becomes a thinking node in the attack network, not just another door."The research, which has not been peer-reviewed yet, was published as governments, security experts and AI companies continue to debate whether generative AI will make sophisticated cyberattacks easier to carry out. One reason the study has attracted attention is that the researchers did not rely on a frontier model from a major AI company, like OpenAI's ChatGPT or Anthropic's Claude. Instead, they used a much smaller open-weight model that can be downloaded and run offline on normal computers.The researchers did not use leading AI models like ChatGPT and Claude. (Image credit: Jaque Silva/NurPhoto via Getty Images)"The researchers employed lightweight open-weight models during their demonstration, which are relatively easy to download, remove guardrail components from, and utilize," Hutchins told Live Science. "By using these types of models, the researchers challenged a long-standing assumption that only advanced/edge-type models present risks related to misuse."Vazdar argued that the work highlights how attackers could increasingly automate tasks that currently require skilled human operators, telling Live Science, "The attacker's marginal cost drops to essentially zero. And you can't patch your way out of it, because it doesn't rely on a single vulnerability class. It reasons. Patch one hole, and it finds another."Could attackers use this AI worm in the real world?Not all experts agree with that assessment, however. Although researchers described the system as capable of targeting a wide range of devices, some cautioned that the demonstration took place in a highly controlled environment designed to showcase the concept."This is at best a lab-based proof of concept in a target-rich test environment," Agee said. The test network contained many intentionally vulnerable systems and lacked active endpoint defenses. "The paper shows that the approach is possible, not necessarily that this attack would work reliably in a normally, or even minimally, defended enterprise network," he added.Any internet-connected device running vulnerable versions of software is theoretically susceptible to being exploited via a similar mechanism. This has been a truism of malicious code for decades.Bob Hutchins, adjunct faculty at Lipscomb UniversityThe worm also generated activity that security teams could potentially detect, he noted, including network scanning, repeated exploitation attempts and privilege-escalation behavior."Even a basic monitoring setup could flag some of that behavior," Agee said.Hutchins likewise warned against overstating the findings. "'Could potentially target almost any device' is technically correct and emotionally misleading," he said. "Any internet-connected device running vulnerable versions of software is theoretically susceptible to being exploited via a similar mechanism. This has been a truism of malicious code for decades."Organizations can still defend themselves by using many of the same measures recommended against conventional cyberattacks, Hutchins added, including prompt patching, strong passwords and multifactor authentication (using multiple forms of identification to log in to systems, like a password sent via text message on top of your password).Even so, experts broadly agree that the study could mark a shift in how malware could operate in the future. Rather than relying on fixed instructions written by human attackers, future malicious software may be able to make many tactical decisions on its own.Related stories'I violated every principle I was given': AI agent deletes company's entire database in 9 seconds, then confessesAI self-replication hacks 'no longer purely theoretical,' study finds —‬ ‪but experts say it's too soon to panicClaude Mythos explained: Is Anthropic's most powerful AI model really too dangerous to release to the public?"The attack is important because it shows that an LLM-based agent can reason through different targets and adapt its approach," Agee said.For Hutchins, the study ultimately represents exactly the kind of work academic researchers should be doing. The study authors "are performing precisely what academia should perform ‪—‬ researching a legitimate threat within a controlled environment before malicious actors begin building it outside of that controlled environment," he said.Whether attackers adopt similar techniques remains to be seen. What the researchers have shown is that a relatively small AI model can already play a meaningful role in planning and directing a cyberattack.

If you simply say the name “Albert Einstein” and ask people what comes to mind, many of them will simply repeat his most famous equation, whether they known what it means or not: E = mc². In plain English, that equation states that the energy of an object is equal to that object’s mass multiplied by the speed of light squared. This simple-seeming mathematical relation holds an enormous amount of physics inside of it, including: The ability to spontaneously create new matter-antimatter pairs of particles as long as their rest mass is less than the amount of available energy you have to try and create them. Whenever a matter-antimatter pair of particles annihilates, that annihilation will produce a photon pair with a specific amount of energy given by the masses of the pair of particles that annihilated. And that every time you have a nuclear reaction, whether fusion or fission, if the mass of the products is less than the mass of the reactants, E = mc² tells you how much energy gets liberated by that reaction. This one equation, E = mc², describes how much energy is inherent to any massive particle at rest, including how much energy it takes to create it and how much energy gets released if you destroy it. But most particles in the Universe aren’t at rest, and plenty of particles don’t have any mass at all. In those cases, E = mc² isn’t the full equation needed to describe your system; it’s only half of the full equation. The other half holds a great number of insights, and shows that both halves are required to make physical sense of what’s going on. The production of matter/antimatter pairs (left) from two photons, or from any colliding pairs of matter and antimatter, is a completely reversible reaction (right), with matter/antimatter annihilating back to two photons in the majority of cases. This creation-and-annihilation process, which obeys E = mc², is the only known way to create and destroy matter or antimatter. If high-energy gamma-rays collide with other particles, there is a chance to produce electron-positron pairs, diminishing the gamma-ray flux observed at great distances. Credit: Dmitri Pogosyan/University of Alberta The reason “rest mass” is such an important concept is because motion — the rate of change of an object’s position over time — isn’t an “absolute” physical property in our Universe. Instead, the key lesson from Einstein’s relativity is that irrespective of what your position is or how your position is changing with time, the laws of physics and the constants of nature, including the all-important speed of light, are always going to appear identical in any inertial (non-accelerating) frame of reference. So if, for example, you have a clock where “one second” is defined by how long it takes light, moving at the speed of light, to: rise from the bottom of the clock to the top, reflect off of a mirror at the top, and come back down to the bottom once again, then two observers in relative motion to one another will experience the passage of time differently. From the perspective of one observer, they’ll be the ones at rest, and their definition of a “second” is the correct one: a round-trip for that light, to go from the bottom to the top to the bottom of the clock, that defines their passage through time. For anyone in motion relative to them, that additional motion means that those external, moving clocks appear to run slow, as their light will appear to require a longer path to rise to the top, reflect, and return to the bottom of their clocks. A light clock will appear to run differently for observers moving at different relative speeds, but this is due to the constancy of the speed of light. Einstein s law of special relativity governs how these time and distance transformations take place between different observers. However, each individual observer will see time pass at the same rate as long as they remain in their own reference frame: one second-per-second, even though when they bring their clocks together after the experiment, they’ll find that they no longer agree. Credit: John D. Norton/University of Pittsburgh The reason for this is that motion through space and time are connected and inextricable: woven together into a fabric known as spacetime. The maximum “motion through time” you can possess is what you experience when you’re at rest with respect to the Universe, which corresponds to when your motion through space is zero. If you do move through space, however, your motion through time must slow in response, which is why the closer you move to the speed of light, the less you age and the slower you experience the passage of time. This has led to a myriad of practical applications, from global positioning systems (GPS) to high-energy particle physics. It’s under these conditions that we’re compelled to look at another part of Einstein’s E = mc²: when you’re in motion, your energy isn’t just given by your rest-mass energy, which is the mc² contribution to your energy. Instead, you also have kinetic energy: the energy that arises due to motion. Whenever two objects collide, whether they stick together (inelastically) or bounce off of one another (elastically), it’s the kinetic energy that they possess, based on their motion relative to one another, that determines how fast they’ll each wind up moving after they crash into each other. This “energy of motion,” or kinetic energy, is essential to the physics of objects in motion, from billiard balls to automobiles to planetary systems. In Newton s theory of gravity, orbits make perfect ellipses when they occur around single, large masses. The presence of other masses, like the other planets, cause these elliptical orbits to precess. However, in general relativity, there is an additional precession effect due to both the curvature of spacetime and the fact that the planets are in motion with respect to the Sun, and this causes the orbit to shift over time, in a fashion that is sometimes measurable. Mercury exhibits the largest such effect within our Solar System, precessing at a rate of an extra 43 (where 1 is 1/3600th of one degree) per century due to this additional effect. Credit: dynamicdiagrams.com, 2011, now defunct However, if you look closely, you’ll notice that Einstein’s most famous equation, E = mc², has absolutely no dependence on motion at all! If energy is simply mass multiplied by the speed of light squared, then how does motion factor into this? Where does the energy of motion for “kinetic energy” even come from? It becomes quite clear that E = mc², alone, can’t do the job. Perhaps an even more compelling argument that there must be more to the story comes if we consider light: a quantum of energy that has no rest mass at all. Light, whether we treat it as a wave whose energy is defined by its wavelength or a particle whose energy is quantized into packets known as photons, has no rest mass, so the m in E = mc² has to equal zero. But light carries energy, so E = mc² can’t be all that there is, or E would equal zero, too, which simply isn’t true. There’s a hint toward the solution if you took physics in High School or college, and learned about the “standard” formula for kinetic energy: KE = ½mv², where v is the speed of the object in motion. This formula only applies at speeds that are low compared to the speed of light: where v is much smaller than c, the speed of light in a vacuum. (That’s the same “c” that appears in E = mc²: 299,792,458 m/s.) This 1934 photograph shows Einstein in front of a blackboard, deriving special relativity for a group of students and onlookers. Although special relativity is now taken for granted, it was revolutionary when Einstein first put it forth, and it doesn’t even describe his most famous equation, which is E = mc², or his most famous advance, which is our current theory of gravitation: general relativity. Credit: public domain The reason that “kinetic energy” offers such a useful hint is because it leads you one step closer to the real key concept in completing Einstein’s most famous equation: momentum. Momentum is the “quantity of motion” that an object has, and it’s well-defined whether the thing that’s in motion is massive or massless, and, if it is massive, whether it moves close to the speed of light or not. Momentum, labeled with a p for very Latin reasons — arising either from the verb “pellere” (to push forcefully) or “petere” (to go) — is basically a measure of how much “Ooomph!” an object has to its motion, and consequently, how difficult it is to bring that moving object to rest. For massive particles moving slow compared to the speed of light, momentum can be well-approximated by the simple formula p = mv. For massive particles moving at any speed, even at a substantial fraction of the speed of light, momentum is more precisely written p = mγv, where “γ” is the Lorentz factor: 1/√(1-(v/c)²). And for massless particles, like light, that move at the speed of light and have no rest mass at all, momentum can’t be written in terms of mass, but can be written in terms of energy very simply, as p = E/c. The particle tracks emanating from a high energy collision at the LHC in 2012 show the creation of many new particles. By building a sophisticated detector around the collision point of relativistic particles, the properties of what occurred and was created at the collision point can be reconstructed, but what’s created is limited by the available energy from the kinetic energy of the colliding particles, with new particles capable of being created from that available energy, limited by Einstein’s E = mc². The maximum LHC energies are nearly a factor of a trillion (10^12) lower than the energies present at the start of the hot Big Bang. Credit: Panos Charitos/Wikimedia Commons user PCharito If we want to give the true expression for the energy inherent to any particle, then, we need to include the effects of its quantity of motion on energy as well as its rest mass’s effects on energy. E = mc², as simple, compact, and notorious as it is, only applies to massive particles at rest: a useful quantity only in certain special cases. Fortunately, there’s an almost-as-simple formula that incorporates both the rest mass energy of a particle, when present, along with the contribution of its quantity of motion to energy as well. That formula for energy is as follows: E = √(m²c⁴ + p²c²). Think about what happens in all the different cases that apply here. If momentum (p) is zero, then that last term goes away entirely, and you simply get E = √(m²c⁴), which just becomes good old E = mc² once again: Einstein’s original rest mass-energy equivalence equation. This is our sanity check: in the original special case, we do indeed get Einstein’s original “most famous” equation back. When a meson, such as a charm-anticharm particle shown here, has its two constituent particles pulled apart by too great an amount, it becomes energetically favorable to rip a new (light) quark/antiquark pair out of the vacuum and create two mesons where there was one before. A strong enough electric field, for long-enough lived mesons, can cause this to occur, with the needed energy for creating more massive particles coming from the underlying electric field, and with the amount of energy required to create these new particles (or particle-antiparticle pairs) described by E = mc². Credit: The Particle Adventure/LBNL/Particle Data Group What if we’re moving slow compared to the speed of light, and we just put in p = mv for momentum? Then the equation becomes E = √(m²c⁴ + m²v²c²), or, if we pull out an mc² from inside the square root, E = mc² * √(1 + (v/c)²). This might not look particularly familiar to you, but consider the following: this equation only works for values of speed, or v, that are slow compared to the speed of light, or c in this equation. Therefore, the part of the equation that reads √(1 + (v/c)²) is only going to be a little bit greater than one, because the (v/c) term is small. Whenever you have, in mathematics, an expression that’s √(1 + x), wherever “x” is small compared to 1, it can be excellently approximated by 1 + ½*x, with the approximation being better for smaller values of “x.” If we apply that to our expression for energy, we turn √(1 + (v/c)²) into 1 + ½*(v/c)², which turns our expression for energy into E = mc² * (1 + ½*(v/c)²), which then becomes, when we multiply the terms out: E = mc² + ½mv². This should make us very happy; it tells us that the total energy is the rest mass energy (the mc² part) plus the kinetic energy (the ½mv² part): the same equations that they teach you in an introductory physics class! This image shows a hole that was made in the panel of NASA’s Solar Max satellite by a micrometeoroid impact. Although this hole likely arose from simply a piece of dust, the “v²” term in the equation for non-relativistic kinetic energy (½mv²) can become very large, very quickly. For particles that move close to the speed of light, the effects of kinetic energy, or energy-of-motion, become even more severe when relativistic effects are taken into account. Credit: NASA When we have a massive particle that moves close to the speed of light, however, the approximation that p = mv can no longer be made with any sort of reliability; that approximation would lead to large errors. Instead, you simply have to calculate the full quantity for yourself, using the full equation, E = √(m²c⁴ + p²c²). As you begin to consider very large momentums, which is precisely the case that we deal with in our largest and most powerful particle accelerators, the rest mass term winds up contributing very little to the overall energy. At 99.999%+ the speed of light, the m²c⁴ term will be much smaller than the p²c² term in the equation, which means we can justify neglecting it. If we do neglect the rest mass term, then we simply get E = √(p²c²), which becomes E = pc: the equation for the energy-momentum relationship for photons and other massless particles. We sometimes call this the ultra-relativistic approximation, as it’s useful wherever the rest mass energy of a system is small compared to the energy due to motion; we can neglect that first term — the m²c⁴ term — even if the object moving ultra-relativistically isn’t truly massless. The longer a photon s wavelength is, the lower in energy it is. But all photons, regardless of wavelength/energy, move at the same speed: the speed of light. This is, surprisingly, irrespective of the motion of the observer relative to light; the speed of all forms of light is measured to always be the same for all observers. Credit: NASA/Sonoma State University/Aurore Simonnet What’s kind of remarkable about this story is the timeline: a key test of Einstein’s general relativity would arrive in 1919, when a total solar eclipse occurred. According to Einstein’s theory, the presence of a large amount of energy, all in one location in spacetime (the Sun), would bend and distort the path of all objects that traveled close to it. This included the light from background stars, which, although massless, would still follow the path created by curved space: the brand-new important concept that came onto the scene with general relativity. But what would the older theory that general relativity was trying to supersede — Newton’s theory of universal gravitation — predict? Some people insisted that it would predict zero deflection, since light had no rest mass and Newton’s theory relied solely on mass for gravitational attraction. But others recognized that photons still carried energy in the form of E = pc, and therefore, if you used the energy that photons had in place of where you would’ve typically used mass (i.e., if you substituted a photon’s E/c² in place of the Newtonian mass, m), you could actually predict a deflection for Newtonian gravity, too, with an assist from Einstein’s 1905 work. The fact that Einstein’s theory of gravitation predicted twice the deflection of even the Einstein-aided Newtonian value, and that general relativity’s predictions were indeed borne out by the observations, provided the key test that enabled us to verify and validate Einstein’s theory, leading to a revolution in how we understood the Universe. Yet the very ability to conduct such a comparison relied on Einstein himself, and the profundity of that simplified version of his most famous equation: E = mc². The results of the 1919 Eddington eclipse expedition showed, conclusively, that the general theory of relativity described the bending of starlight around massive objects, overthrowing the Newtonian picture. This was the first observational confirmation of Einstein’s theory of gravity. Credit: London Illustrated News, 1919 When you think of Einstein’s most famous equation, you should still recognize how profound the simple statement, E = mc², actually is. It tells us that every massive particle has an inherent amount of energy inherent to it, even when it’s at rest, and that its energy can never drop below that key value: mc². If you want to create a particle like it, you require at least that much energy; if you must create that particle along with its antiparticle counterpart, you require at least double that energy. And if you destroy or annihilate any massive particle away, all of that rest mass energy, all mc² of it, will become part of the energy that all of the “daughter particles,” or particles produced in the annihilation, carry away. But you should also recognize that E = mc² is only half of the full story: because particles exist not merely at rest, but they also move through the Universe. The quantity of motion that they carry with them, momentum, leads to a certain amount of energy-of-motion being associated with that particle as well. For slow-moving, massive particles, you can approximate that energy-of-motion with E = ½mv². For massless particles and ultra-relativistic massive particles, you can approximate that energy of motion with E = pc. But if you want the general case, where rest mass and momentum both are included, you need the full equation for the energy of a particle: E = √(m²c⁴ + p²c²). As famous as it is, E = mc² is only half of the full equation that’s needed to describe a particle’s energy. To get the other half, you have to remember that you cannot simply describe the Universe by taking a snapshot of it. It has the kind of beauty — and energy — that can only be captured if motion is included as well. This article was first published in May of 2023. It was updated in June of 2026. This article Einstein’s E = mc² is only half of the full equation is featured on Big Think.

Operational performance is becoming just as important as investment performance in private credit. As fundraising slows and investor expectations increase, firms are facing growing pressure to modernize the IT infrastructure supporting their portfolios.In fact, transparency and faster reporting are becoming top priorities.Without such capabilities, funds can’t clearly see across their own portfolio, amplifying stress when markets are less forgiving. Especially as private credit scales and operational diligence become more central to allocation decisions, such back-office issues become something more structural. Fortunately, managers that properly invest in their operational foundations now, will be better equipped to manage the increasing demands facing the private credit industry moving forward. Analyzing Operational PressuresPrivate credit is operationally intensive. And many firms never built systems to match the increasing complexity of their portfolios as they grew and evolved. Instead, operations are often spread across legacy servicing platforms, spreadsheets, email-driven workflows, and disconnected internal tools, leaving firms without a single real-time view of portfolio data.Most loan servicing operates on cycles such as monthly reconciliations, quarterly reporting, and batch-based payment processing. This model reflects the constraints of legacy systems and manual workflows. Data must be collected, validated, and processed in stages. Consequently, funds often view their portfolios through periodic snapshots rather than in real-time.Many funds also maintain parallel spreadsheets to verify their servicer's calculations. Known as shadow booking, this redundancy has no other purpose than to increase control. Ultimately though, it’s a sign of mistrust in the data provided, and the underlying calculations are often difficult to inspect. When discrepancies arise, they are discovered after the fact and require manual investigation, often across multiple systems.These realities of the private credit industry are now colliding with a more competitive fundraising environment. Even if their underlying deals aren't catastrophic, private credit funds begin to look fragile if they cannot perform operationally well under increasing pressure. Not surprisingly, many private credit firms are looking to AI to address these issues. But automation software alone cannot repair broken operational foundations. The critical question becomes whether their infrastructure is actually prepared to support the AI model they choose to use.AI Alone Is Not the AnswerAI agents can execute operational work reliably while platforms can provide real-time visibility and full auditability. The pieces are in place. But there’s a pattern in how AI projects fail in the finance industry that's worth naming and it's almost never the model that's the problem.What’s often missing is the infrastructure surrounding the model - the systems, workflows, data access, permissions, and controls that allow AI to operate reliably inside real financial processes. Otherwise known as the harness. Generic AI tools don’t know what a rate notice is. They don't know that a prepayment request triggers a multi-step workflow across every syndicated entity on a facility. They don't know the difference between a funded tranche and a committed-but-undisbursed revolver, or why that distinction matters for how a payoff figure should be calculated.This lack of operational context is also one of the biggest reasons AI hallucinations occur in finance. The majority of hallucinations aren't random malfunctions, they're a context problem. The AI model wants to give an appropriate answer. When it doesn't have access to the specific data it needs, it reasons from what it does know and produces something plausible. Private credit portfolio data isn't embedded in any public language model. If the harness doesn't provide it, through tools, memory or real-time data access, the model will fill the gap with something that sounds right. Which, in a financial operations context, is a real problem.The fix isn’t a better model. It’s a harness that gives the agent access to the right data, at the right moment, with the right tools and controls to retrieve it.The firms that focus on building operational systems that provide context, transparency, audibility and human oversight will receive the greatest value from its AI investments. In private credit, the long-term advantage may not come from adopting AI faster than competitors, but from building the infrastructure, or the harness, capable of supporting it responsibly and at scale.Establishing New Competitive DifferentiatorsAs these operational systems mature and advance, they will also reshape what excellence actually looks like inside private credit firms. Responsiveness won't be a differentiator. It will be assumed. Real-time and instant delivery will be the new baseline.This is because most routine interactions will no longer require human involvement. With real-time systems, shared data layers, and automated workflows, information will be directly accessible and continuously updated. What previously required a request and response cycle will be resolved at the source.As a result, the role of the servicer shifts. They are no longer measured by how quickly they process or reply, but by how well they handle what cannot be automated - exceptions, edge cases, and judgment calls.This is why the next generation of private credit leaders may look fundamentally different from the firms that defined the industry’s earlier growth period. Capital access and underwriting expertise will remain essential, but operational execution is becoming increasingly strategic.Most funds are using general-purpose AI for ad-hoc analysis or are in a holding pattern. A small number are starting to build their own and discovering how much harder it is than first expected. The funds that move first on specialized operational infrastructure will have an advantage that compounds. Not because they picked the right model (the model will keep getting better and cheaper regardless), but because they built or adopted the right harness, trained it on the right context, and gave it the controls that make it trustworthy at scale.In many ways, private credit firms are evolving into operational organizations as much as financial organizations. The ability to manage workflows, data, oversight, and execution will become a defining part of a firm’s competitive performance.We feature the best personal finance software.This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit