LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
TodoTecnologíaProductosInversiónBolsaOtros

Lectura

2026-11-17124 publicaciones

Actualizar

Google Adds Interactive Site Generation to Gemini Notebook

Material
2026-11-17 00:00 GMT+8

U.S. tech giant Google announced on November 16, 2026, that its AI note-taking and research tool, Gemini Notebook, has added a new core feature: users can now type “make me a site” in the chat along with a description of what they want to build, and the tool will transform uploaded notes, research, or sources into interactive websites. This update aims to help users quickly make sense of complex information.

According to the official Google Blog post, the generated interactive sites include various types, such as visual explainers for exam preparation (like an interactive periodic table or historical timeline), interactive dashboards for exploring data (such as quarterly sales performance), and project management trackers for outlining tasks and deadlines. Once built, users can select “Open in a new tab” to interact with the site or share it with others.

This capability is rolling out globally on the web today for all Gemini Notebook users (18+), including Google Workspace subscribers, with mobile app support coming soon. This move marks a shift in AI note tools from simple text Q&A and summarization towards generating dynamic visualization tools, lowering the barrier for users to create data dashboards and educational aids.

Fuentes:Google

Investigación

Scope Creep in AI Agents: CodeRabbit Founder Harjot Gill and OpenCode’s Ryan Vogel on Why Deciding What to Keep Is Harder Than Generation

Tema · AI代理代码范围蔓延Material
2026-11-17 04:00 GMT+8

When AI agents can produce two hundred pull requests (PRs) a day or pile up six hundred pending changes for review at once, the core issue facing engineering teams is no longer whether the code has bugs, but whether these changes should exist at all. CodeRabbit founder Harjot Gill and OpenCode maintainer Ryan Vogel discussed this issue on The Merge podcast. Their consensus is that while AI-written code is useful, the harder part is deciding what to merge, what to push back on, and what to drop.

This judgment problem has an easily overlooked starting point: the problem takes shape before the code exists. Vogel gave the example of a kitchen to-do app—assign the task to an agent, and it will first add a database, then a caching layer, then more infrastructure. Each step individually does not look wrong, but the application ends up far more complex than a simple to-do list requires. Vogel pointed out that without an external reference point, every decision provides justification for the next, and the project can drift off course even while each step appears productive.

To combat this creep, Vogel and Gill each approached from different stages. Vogel’s method involves having the agent output an architecture plan first, which humans then challenge by identifying unnecessary parts; simultaneously, he runs a separate agent to compare intermediate outputs against the original requirement direction. Gill, drawing from CodeRabbit’s practice, added another dimension: review should not just look at diffs, but read changes in the order they were built. CodeRabbit divides review into three layers—the bottom layer consists of data and schema changes, the middle layer covers functions, business logic, and API routes, and the top layer includes the interface and its tests. Schema changes explain why an API behaves differently, and API changes explain what the interface now needs to handle. Reviewing in this order allows humans to check the assumptions connecting various parts, rather than guessing from filenames.

But sequence alone is not enough; evidence is also required. Gill mentioned hallucination issues encountered early on by CodeRabbit: when models are forced to give answers, they fabricate them. The team’s solution was to provide an escape hatch—for situations with insufficient information, the model could ask questions, investigate further, or write a script to verify its hypotheses. This reduced the hallucination rate. In the review process, the same discipline means treating a suspected bug as something to be tested, rather than immediately turning it into a comment requiring developer action. The recipient of the review feedback can inspect how the conclusion was reached, run the check, and question its assumptions. Scripts themselves may still be incorrect or incomplete, but their value lies in making part of the claim verifiable.

This approach has a clear boundary of applicability. Gill pointed out that agents are best suited for tasks with clear completion conditions, such as adapting a working integration to another platform—because both the agent and the reviewer have concrete results to check. Open-ended design work requires more frequent human decision-making because the desired outcome is still taking shape while the agent runs. In other words, the premise for reference points to be effective is that you already know where you are going; if the destination itself is changing, the reference point loses its anchoring function.

Fuentes:CodeRabbit

Investigación

NVIDIA Releases cuPotential Library to Accelerate Molecular Dynamics Simulations by Up to 3x

Tema · cuPotentialResumen rápido
2026-11-17 06:30 GMT+8

NVIDIA released cuPotential on November 16, 2026, a modular C API library designed specifically for molecular dynamics (MD) simulations. The library addresses the computational bottleneck of evaluating long-range interactions, such as electrostatics and dispersion, under periodic boundary conditions in classical force fields. NVIDIA also provided a plugin for OpenMM, allowing users to replace existing nonbonded forces, Langevin integrators, and Monte Carlo barostats with drop-in replacements requiring minimal code changes.

According to data disclosed in NVIDIA's technical blog, cuPotential achieves end-to-end speedups of 2x to over 3x for systems ranging from 24,000 to 400,000 atoms across various GPU architectures, from T4 to RTX PRO 6000 Blackwell. Notably, newer Blackwell architecture GPUs demonstrated speedups exceeding 3x in selected benchmarks. This performance gain is driven by newly developed C2 continuous splitting and spreading functions, created in collaboration with D. E. Shaw Research and the Flatiron Institute, which achieve target accuracy at lower grid resolutions and reduced cutoffs compared to standard Ewald and u-series methods.

Developers can access examples, benchmarks, and integration code via the GitHub repository. The library currently supports Linux x86-64 environments and requires an NVIDIA driver compatible with CUDA 13.1. For OpenMM users, the integration process involves building the runtime and plugin, followed by installing the Python wheel package.

Fuentes:NVIDIA Technical Blog

Investigación

Voice AI Firm ElevenLabs Announces Acquisition of Japanese Lab Kotoba to Expand Market Presence

Tema · ElevenLabsMaterial
2026-11-16 20:00 GMT+8

US-based voice AI company ElevenLabs announced on November 16, 2026, that it will acquire Kotoba Technologies, Inc., a leading Japanese voice AI lab. The acquisition aims to accelerate the adoption of AI agents and voice interfaces in Japan, the world’s fourth-largest economy. The deal is expected to close later in Q4 2026, subject to customary regulatory approvals and closing conditions.

Kotoba was founded in 2023 by Noriyuki Kojima and Jungo Kasai with the goal of developing voice AI for a multilingual world. Kojima holds a PhD in computer science from Cornell University and co-built Japan’s Fugaku-LLM, an open-source model trained on the RIKEN Fugaku supercomputer. Kasai, a PhD holder from the University of Washington and former assistant professor at Toyota Technological Institute at Chicago, is one of Japan’s most-cited researchers. Kotoba’s voice model, Koto, excels in Japanese text-to-speech, speech-to-text, and simultaneous interpretation, and is already used by major enterprises such as Salesforce for Agentforce Voice.

Mati Staniszewski, co-founder and CEO of ElevenLabs, stated that the language people speak and the social norms they follow need to be reflected in technology’s architecture from the start. By integrating Kotoba’s deep understanding of Japanese cultural nuances—such as 'reading the air'—ElevenLabs plans to build local sales leadership and invest in infrastructure in Japan to create more natural and intuitive AI interactions.

Fuentes:ElevenLabs

Investigación

CoreWeave Says Its ML Model with Aston Martin Cuts F1 Wind Tunnel Prediction Error by 50%

Material
2026-11-17 06:53 GMT+8

Cloud computing company CoreWeave announced that a machine learning model developed in partnership with the British Formula 1 (F1) team Aston Martin Aramco has reduced the error in predicting on-track aerodynamic performance by 50%. The model aims to bridge the "delta" between idealized wind tunnel data and actual track conditions, allowing engineers to calibrate car setups more accurately before Friday practice begins.

In F1, wind tunnels provide controlled environments, but dynamic factors on the track—such as speed changes, braking loads, cornering forces, curb heights, and dirty air—cause downforce and drag to vary significantly. Traditionally, teams spent precious Friday practice time discovering and correcting these prediction gaps. CoreWeave's Physical AI Field Engineering team worked with Aston Martin's Vehicle Science team for six weeks, jointly performing feature engineering to extract meaningful patterns from limited, noisy telemetry data collected during race weekends.

According to CoreWeave's official blog, the primary challenge was the small, noisy dataset: although telemetry streams at 100Hz, downsampling and denoising leave very little usable data per weekend. The team explored statistical methods and classical modeling techniques, finding that a novel modern modeling approach outperformed existing solutions. The model was delivered ahead of the 2025 Abu Dhabi season finale and has been used throughout the 2026 season to guide component choices and setup optimization in the week before each race.

Enrico Cardile, Chief Technical Officer of Aston Martin Aramco, stated that while the wind tunnel remains the reference, the track is where performance is decided; the new model allows the team to arrive at each circuit with a clearer picture of how the car will behave, freeing up practice time for refinement rather than discovery. This case illustrates the potential of AI in physical engineering, showing how collaboration between domain experts and algorithm engineers can improve predictions of physical reality in data-scarce scenarios.

Fuentes:wf.coreweave.com

Investigación

China's LineShine Retains Top Supercomputer Crown as AI Data Centers Drive Lowest New System Count in 33 Years

Tema · Top500超算榜单Material
2026-11-17 06:17 GMT+8

According to the 68th edition of the Top500 list published at SC26 in Chicago, China's LineShine, installed at the National Supercomputer Center in Shenzhen, remains the world's highest-performing supercomputer. The machine relies on the homegrown LingKun platform and achieves sustained FP64 compute performance of 2.198 FP64 ExaFLOPS using exclusively CPUs. Despite unprecedented demand for high-performance computing hardware, only 27 new supercomputers made it onto the November list, the lowest number of additions since the project was founded in 1993.

This anomaly reflects a shift in infrastructure priorities. Estimates by Epoch AI cited by Top500 authors indicate that global AI compute capacity is growing by 2.5X per year. As of Q4 2025, five major American hyperscalers accounted for an estimated 71% of the world's AI compute resources. These vast clusters are deployed for commercial AI workloads rather than traditional scientific supercomputing, meaning much of the new hardware never appears in the Top500 ranking.

Neither the top 10 performing supercomputers nor the top 10 most energy-efficient machines changed positions since June. LineShine gains only 3.6X performance when moving from FP64 to mixed precision, lower than systems using accelerators. In contrast, US exascale systems like El Capitan, Frontier, and Aurora rely on AMD or Intel GPU accelerators. Germany's Jupiter Booster remains the most energy-efficient exascale system on the list.

While new entrants were scarce, the aggregate HPL performance of the top 500 machines increased by 4.4% to 19.57 FP64 ExaFLOPS. The entry threshold rose to 2.857 PFLOPS. The number of supercomputers equipped with compute accelerators increased from 275 to 291 (58.2%), indicating that while traditional supercomputer construction has slowed, the integration of AI-specific hardware continues to reshape the high-performance computing landscape.

Fuentes:Tom's Hardware

Investigación

Vanguard CEO Salim Ramji: The Vast Majority of Investment Products Fail the Efficiency Frontier Test of Index Portfolios

Material
2026-11-16 21:30 GMT+8

Salim Ramji, CEO of the Vanguard Group (a U.S.-based asset management company), explained in a podcast interview the core mechanism used internally to screen all new products: an efficiency frontier test colloquially known as "Does it load?" The framework operates straightforwardly: regardless of whether the candidate product is an index fund, an actively managed fund, fixed income, equities, public market instruments, or private equity, it must be compared against the efficient frontier of a simple index portfolio (with allocations such as 60/40, 70/30, or 80/20) to assess its potential to deliver returns superior to that frontier. Ramji’s conclusion was clear: the vast majority of products fail this test.

The reader value of this mechanism lies in how it reduces the long-standing debate among individual investors regarding "active management versus index investing" back to a more fundamental question: it is not a debate of style, but of price. Ramji directly quoted Jack Bogle’s famous adage—"In investing, you get what you don’t pay for"—and pointed out that this principle applies equally to both active and index strategies. He specifically noted that low-fee active funds often outperform high-fee active funds. What ultimately determines results is not the management style, but how much one pays for that style.

Ramji extended this logic to the private markets. He mentioned that Vanguard launched a private market product in collaboration with HarborVest approximately six years ago, and is currently exploring potential collaborations with other institutions such as Blackstone, provided these products pass the same "Does it load?" test. He acknowledged that fee structures in the private market for retail investors are "far above reasonable levels," stating that Vanguard’s approach is to negotiate institutional-grade pricing and increase transparency; otherwise, they will not participate. Regarding the allocation of private assets within 401(k) retirement plans in the United States, his assessment is that while private markets may become one of the available options, they are unlikely to achieve the widespread adoption seen with index target-date funds, because for most people, simplicity itself provides a sense of security.

A necessary caveat is that this framework comes from the CEO of Vanguard, an institution with an obvious commercial interest in promoting index products. Ramji proactively addressed this point in the interview, noting that Vanguard has operated both active and index funds since 1975, with its active fixed-income products dating back to 1981, and that these active products are subject to the same efficiency frontier test. He cited his personal experience as validation of the value of financial advice (after his father passed away, his mother’s advisor helped prevent her from making the costly mistake of converting all assets into cash), indicating that he does not equate "simplicity" with "the need for no human assistance." However, the assertion that "the vast majority of products fail" was not supported by independently verifiable quantitative data in the interview. Readers should treat this as an attributional statement from a leader managing approximately $12.5 trillion in assets, rather than an empirically peer-reviewed conclusion.

Fuentes:The Big Picture

Investigación

ByteDance's Seedance 2.5 Flash Video Model Now Available on Vercel AI Gateway

Resumen rápido
Verificado 2026-11-17 06:44 GMT+8

Cloud development platform Vercel announced that its AI Gateway now supports ByteDance's video generation model, Seedance 2.5 Flash. The model is described as a faster, lower-cost version of Seedance 2.5, with output quality close to the standard model in terms of human realism, reference-based editing, and prompt adherence.

Through this integration, developers can generate videos with synchronized audio from text prompts or image, video, and audio references. Clips can range from 4 to 30 seconds long at up to 720p resolution and 24 frames per second, in landscape, square, or portrait formats. Users can also edit or extend existing videos, or provide images to set the first and last frames.

Developers can call the bytedance/seedance-2.5-flash model using AI SDK 7. To guide generation, URLs for reference images and videos can be passed in the inputReferences parameter, where they are numbered separately in the prompt as [Image 1], [Video 1], etc.

Fuentes:Vercel

Investigación

Theory Ventures Partner Tomasz Tunguz: The Enterprise AI Market is Splitting by Tier—Is It a Pricing Strategy or a Structural Cost Difference?

Material
2026-11-16 08:00 GMT+8

Over 16 weeks, Databricks Unity Gateway collected AI token usage telemetry from more than 10,000 enterprises. Tomasz Tunguz distilled from this data a clear pattern of tiered market splitting: Anthropic captured 94% of the token share in the frontier SOTA (state-of-the-art) tier, while OpenAI held 81% in the low-end segment. Each lab approaches monopoly levels within its own specific niche, but those niches are entirely different.

Anthropic’s SOTA share rose from 81% in July to 94% on November 2, driven by the September 21 release of Claude Opus 5.5. In the SOTA tier, Anthropic’s frontier models (Opus 4.8, Opus 5, Opus 5.5, and Fable 5.1) accounted for 39.8% of total proprietary tokens, whereas OpenAI’s frontier models (GPT 6 Astra and GPT 5.6 Sol) accounted for only 2.5%. This is not a gradual advantage but a gap of magnitude.

OpenAI took the opposite path. On July 30, it cut Luna pricing by 80%, followed by another 50% reduction several months later, for a cumulative drop of 90%. On September 22, GPT-6 Luna launched at $0.10 per million tokens. Consequently, its low-end share surged from 30% to 66%, peaked at 84%, and eventually stabilized at 81%. Meanwhile, the share of open-weight models climbed from 7% to 33% between July and September, but stalled at 31% in the same week that Luna and Sol prices were cut by 50%—suggesting that price sensitivity is real.

The mid-tier is where the two labs collide head-on. OpenAI’s Sol series went from holding 0% of the blended Sonnet and Sol mid-tier share in July to 47% by November 2. In the latest queue of Sonnet 5.5 versus Sol 6/6.1, it captured 57% within six weeks.

Tunguz’s core question is not about who won how much share, but what caused this tiered landscape. One explanation is a deliberate pricing strategy: Anthropic chooses the high end, OpenAI chooses the low end, and each takes what it needs. Another explanation is structural cost differences: the model architectures of the two labs differ fundamentally in the cost of serving tokens, giving each natural competitiveness in different price ranges.

The implication for investors and enterprises is significant: if it is a strategy, the share landscape can be altered by pricing actions, and competition intensity depends on rivals' responses; if it is structural, the landscape is more durable, entry barriers are higher, and the meaning of share changes is entirely different. Tunguz does not provide a definitive answer, but he outlines a testing path—whether Luna’s 90% price cut was a short-term market-share subsidy or a structural cost advantage in architecture will be revealed by share trends in the coming weeks.

It is important to note that Databricks Unity Gateway reflects the usage patterns of its enterprise customer base, primarily data platforms and agentic coding workloads, and cannot be directly equated with the full picture of the entire enterprise AI market. Share data measures token usage volume, not revenue; the figure that Anthropic accounts for 70% of proprietary spend comes from the same dataset, but the relationship between token share and dollar share depends on price differences across tiers.

Fuentes:Tomasz Tunguz

Investigación

Ollama and Together AI Joint Article: Cache Invalidation in Long-Horizon Agent Sessions Almost All Stem from Violating a Single Prefix-Matching Rule

Tema · Agent前缀缓存失效Material
2026-11-16

Every step an agent takes, the harness resends the entire conversation to the model: the system prompt, tool definitions, user requests, every file the agent has read, and every command it has executed. What typically returns is only a few dozen tokens of tool calls. This is the typical traffic pattern for agent workloads—extremely long input sequences with very short outputs. If the GPU has to process the entire history from scratch on every turn, the longer the session, the slower and more expensive it becomes.

Prefix caching is the mechanism that prevents this degradation, but its rules are extremely simple: the cache matches from the beginning of the prompt and stops at the first differing token, after which all content must be reprocessed. Appending new content to the end of the prompt incurs no extra cost; changing any character earlier, even just one, means everything after it must be recalculated. If the change occurs at the very top, it effectively invalidates the entire session. Parth Sareen of Ollama and Zain Hasan of Together AI point out in their joint article that almost all agent cache issues derive from this single rule.

The easiest mistake on the harness side is placing content that changes every turn into the system prompt. Timestamps, git status, or the working directory the agent is currently switching through—if the harness re-renders the system prompt with fresh values on every turn, every turn becomes a full re-read. The fix is to place such information in the task message, write it once, and keep the system prompt completely identical across all sessions.

A more subtle example comes from GLM-5.3-Flash: reasoning effort is not a setting handled independently by the engine; the chat template writes it into the very first system block, ahead of tools and conversation. When a user adjusts effort from low to high, the next request differs from the previous one almost from the first token, causing the entire session to be reprocessed. In long tasks, this equates to spending hundreds of thousands of tokens just to change one setting. The authors' advice is to select an effort level and stick with it; if you truly need more thinking, starting a new session is often cheaper than continuing the old one, and you also get a clean context.

Tool lazy-loading is another decision made on the harness side. Modern harnesses support agents searching for and loading tool definitions at runtime. If new definitions are added to the tool list—the tool list sits at the top of the prompt, before the conversation—the prefix changes, and everything after it is recalculated. A better approach is to place definitions at the end, since new content is naturally appended there. However, not all models support inserting tool definitions mid-conversation; some chat templates will simply error out. Harness developers need to know this; users do not.

Large tool results also require care. An agent reads a 640-line file, but the tool returns only the first 100 lines without indicating truncation. The model does not know it is missing data and may confidently edit code it has never seen. The fix requires just one line of output: telling the model it received lines 1 to 100 of 640, and that it can search or request other ranges. Truncating saves tokens, but truncating without explanation leads to more costly errors by the model.

Fuentes:together.ai

Investigación

Divergence in Korean Game Giants' Q3 Earnings: Legacy IPs Sustain Profits While Release Delays Hurt Performance

Tema · 韩国游戏三季度业绩分化Material
2026-11-16 21:47 GMT+8

Major South Korean game companies delivered mixed results for the third quarter of 2026. According to a report by Byline Network, the presence of verified Intellectual Property (IP), the timing of new releases, and cost management capabilities collectively determined each company's bottom line. Notably, an increase in revenue scale did not necessarily translate into improved profitability, with several firms falling into a trap of 'revenue growth without profit growth' due to surging costs.

Companies with strong existing IP portfolios performed steadily. Nexon achieved record quarterly sales driven by the 'MapleStory' franchise and 'FC Online', reporting revenue of 1.23 trillion KRW (up 17% YoY) and operating profit of 371.8 billion KRW (up 11% YoY). NCSoft saw its revenue surge 91% YoY to 686.5 billion KRW and turned profitable with 119.6 billion KRW in operating profit, thanks to leveraging legacy IPs like 'Lineage Classic' and diversifying into mobile casual games. This indicates that deepening user value in existing titles and expanding lightweight categories can effectively smooth performance fluctuations when blockbuster new releases are absent.

The success or failure of new releases directly led to polarized results. Com2uS reported a 35.2% YoY revenue increase to 178 billion KRW and returned to profitability, driven by its new MMORPG 'Zeus: God of Pride' topping app store charts shortly after launch. Devsisters also achieved a 47.9% revenue growth and turned profitable, aided by the overseas popularity of new titles like 'Cookie Run: Crumble'. Conversely, Netmarble saw revenue drop 8.8% YoY to 634.7 billion KRW and operating profit shrink 44.6%, primarily because key new titles were delayed until next year. Kakao Games recorded a 46% revenue decline and an operating loss of 26.3 billion KRW due to a gap in new releases and declining sales of older titles.

Cost management emerged as a critical variable distinguishing quality earnings from mediocre ones. Although Krafton hit cumulative revenue highs, its Q3 operating costs surged approximately 43% YoY, largely due to the consolidation of Japanese ad firm ADK Group and increased marketing expenses. This resulted in a 35.5% YoY drop in operating profit to 225 billion KRW, despite an 11.6% revenue increase. Pearl Abyss faced a similar scenario with revenue up 18% but operating profit down 8.8%, as labor and outsourcing costs rose faster than income. This warns investors to be vigilant about how M&A integration costs and marketing spend erode short-term margins during industry expansion phases.

Furthermore, non-game businesses provided a hedge. Wemade's game revenue fell 11%, but its operating profit edged up 0.3% thanks to the recognition of approximately 50 billion KRW in licensing fees for 'Legend of Mir 2/3' in China. Com2uS Holdings turned profitable through improved investment gains from affiliates and cost cuts. Looking ahead to Q4 and next year, Netmarble plans to launch over 13 new titles, and NCSoft has raised its full-year revenue guidance to 2.8 trillion KRW. The industry's competitive focus is shifting from pure traffic acquisition to the efficiency of long-term IP operations and the execution of new release pipelines.

Fuentes:바이라인네트워크

Investigación

US BIS Issues New Rule Restricting Remote Access to AI Servers to Close Export Control Loophole

Tema · BISMaterial
2026-11-17 05:50 GMT+8

The U.S. Commerce Department’s Bureau of Industry and Security (BIS) released a new Interim Final Rule (IFR) on November 16, 2026, designed to ensure that physical export restrictions on advanced artificial intelligence hardware cannot be bypassed through cloud-based architectures. Developed in collaboration with the Department of Treasury and the National Security Council, the rule aims to prevent companies in non-whitelisted countries from accessing restricted U.S. technology by renting data center capacity in third-party nations.

Under the new regulations, overseas data centers will be required to take more steps to limit access to their services where U.S.-based technology is being used. Restrictions will specifically target remote access provided to companies headquartered in higher-risk countries, such as China, Russia, and Iran. Additionally, licenses will be required for anyone attempting to export advanced chips to companies headquartered in Indonesia, Malaysia, Singapore, Thailand, or Brazil. These requirements are set to officially go into effect in 90 days.

This action follows reports that existing limitations on hardware exports were being circumvented. While Nvidia is only allowed to export its high-performance AI chips to whitelisted companies, there have been ongoing claims of black market GPU sales and crackdowns on overseas AI GPU rentals from China. For instance, Nvidia Blackwell Ultra B300 GPUs were recently discovered in a data center used by China Telecom. Although Nvidia had previously told data center operators to limit remote access—requiring that 51% of their business come from non-Chinese companies—this voluntary approach was deemed insufficient by regulators, prompting this stricter legal intervention.

Fuentes:Tom's Hardware

Investigación
Siguiente →
Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión

Finally, both Gill and Vogel emphasized the handoff issue. The handoff from agent to developer receives most attention, but the handoff between people is equally important. Code alone does not show why a certain approach was chosen, what was tried and abandoned, which choices were deliberate, and which were merely defaults. Vogel’s team shares the full session logs behind changes, allowing colleagues reviewing the code to also see the prompts and the decisions that produced them. Gill mentioned CodeRabbit’s collaborative memory, where one developer’s review feedback influences subsequent reviews. Both methods allow the next person to see what was attempted and what still needs validation. Not every change deserves the same level of scrutiny—a small interface change that can be rolled back in minutes carries completely different consequences compared to a change on a mature code path relied upon by many people, and review effort should reflect this difference.

How sub-agents are launched determines how much of the parent session's cache they can reuse. Inheriting the entire parent session (system prompt, tools, conversation history, plus the new task) means the prefix is already cached, resulting in low processing costs; sending a brief (same system prompt and tools, but only providing necessary files and instructions) means the cache covers only the setup phase. Inheritance sounds strictly better, but a cached prefix is cheap, not free—every token generated by the sub-agent still has to read through it, and it occupies GPU memory throughout. A reviewer needing only two files and a checklist likely does not need three hundred turns of history.

The engine's responsibility is to ensure the cache persists when the next request arrives and that the request is routed to the machine holding the cache. Together AI describes two routing paths: the client sends a prompt_cache_key, and requests with the same key are routed to the same node; or the prefix is hashed, and the request is sent to the node with the longest matching cached prefix. The latter is another reason why system prompts must remain consistent across sessions—if every new session starts with exactly the same system prompt and tool definitions, a new session might directly hit an existing cache.

However, routing does not unconditionally favor warm nodes. A warm node may be overloaded, requiring a trade-off between waiting in queue and going to an idle node to reprocess the prefix. The authors explicitly state that under load-aware routing paradigms, a cache miss is sometimes the more cost-effective choice. This means that even with a perfectly stable prompt, a cache hit is not guaranteed—the quality of the backend engine implementation is also a constraint.

Cache survival time is also an issue. After an agent makes a tool call, it may wait minutes for tests to run. Keeping this idle cache in GPU memory occupies space needed by active users; discarding it means the agent must re-read everything upon return. Together AI's approach is to offload the cache of idle sessions to CPU memory and restore it for the next turn.

Viewing both sides together: the harness decides where the first change lands, and the engine decides whether the cache is still there. Getting it right on either side alone is insufficient. If the harness keeps the prompt perfectly stable but the engine sends turn 7 to the wrong machine, the cache is lost. If the engine routes perfectly but the harness changes the first line every turn, the cache is still lost. The speed an agent user sees at turn 90 depends on both sides simultaneously adhering to that prefix-matching rule.