LeituraPesquisaRadarFramework de investimento
Entrar / Cadastrar
TudoTecnologiaProdutosInvestimentoBolsaOutros

Leitura

2026-11-10115 posts

Atualizar

US Certificate Authority DigiCert Confirms Data Breach Exposing Customer Identity Documents

Tópico · DigiCertMaterial
2026-11-10 09:46 GMT+8

US-based certificate management company DigiCert confirmed on November 5 that a data breach occurred due to a configuration error during a server migration in February 2025. This error inadvertently exposed an internal business system, originally accessible only via VPN, to the internet. Attackers exploited an authorization control vulnerability to create unauthorized accounts and gain read-only access to archived customer data.

According to notifications sent to affected customers, the leaked information includes names, corporate email addresses, phone numbers, and company addresses. Because some customers had previously provided identity verification documents to complete certificate validation procedures, the breach may also have exposed driver's licenses, passports, utility bills, and other documents containing names and addresses.

In an incident report submitted to Mozilla Bugzilla, DigiCert stated that it first discovered unauthorized accounts in the system in June 2026 but initially believed they could not access customer information because they lacked application assignments. It was only after receiving reports through its security research program that the company confirmed external actors could create new accounts. Further investigation on October 19 confirmed that unauthorized accounts had accessed customer data between June 27 and July 2, 2026.

Fontes:iThome

Pesquisa

South Korean Physical AI Firm Ndotlight Supplies KT with Simulation Data for Flexible Objects

Tópico · NdotlightVisão rápida
2026-11-09 14:42 GMT+8

South Korean physical AI data company Ndotlight has announced that it is supplying KT, the country's leading telecommunications operator, with 3D simulation data for flexible objects such as cables. This partnership aims to address a key bottleneck in industrial robot deployment: accurately simulating the bending and stretching properties of cables in virtual spaces so that robots can practice and validate tasks before entering actual work sites.

Ndotlight generates these "SimReady" assets using its self-developed CAD platform, TRINIX. The platform handles the entire process from 3D model creation to inputting physical attributes like weight and elasticity, as well as joint information. The output conforms to the OpenUSD standard, allowing direct integration into Nvidia's Isaac Sim robot simulation environment for building and verifying robot operation scenarios.

The announcement was made during "Nvidia AI Day Seoul 2026." Ndotlight CTO Kim Seon-tae presented the case study with KT and showcased "DexBench," a robot hand manipulation evaluation project involving Ndotlight, Nvidia, and RealWorld. Ndotlight stated plans to contribute its data to Nvidia's "SimReady Central" repository to expand into global markets.

Fontes:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Pesquisa

Anthropic Blog Author Addy Osmani: Select the Claude Model by Adjusting Effort Levels First, Then Switching Models

Material
2026-11-09

In an official blog post released on November 9, Anthropic provided a clear starting recommendation: most coding and knowledge work should begin with Opus 5.5 at its default effort level. This is not because Opus 5.5 is the strongest model—Fable 5.2 holds that title—but because it sits at the optimal balance point between cost and capability. Opus 5.5 costs $4 per million input tokens and $20 per million output tokens, exactly half the price of Fable 5.2 ($8 and $40 respectively).

However, what makes this blog truly noteworthy is not the starting recommendation, but the adjustment sequence it proposes. Author Addy Osmani writes that you should adjust the effort level before switching models. Effort is a finer dial than model switching, simultaneously controlling changes in both cost and capability. Beyond the high setting, moving up one more effort level may cost as much as switching to a larger model. Conversely, dropping down one level might yield the results you need at a lower rate.

This advice is backed by specific benchmark data. On Anthropic's self-tested Terminal-Bench 3.0, the previous generation Fable 5.1 scored 58.0% at max effort, while Opus 5.5 scored 58.9% at high effort using only half the number of tokens. On FrontierCode 2.0, Fable 5.2 scored 45.3% at high effort, essentially matching Opus 5.5's 45.5% at max effort, but at roughly half the cost per task. These figures demonstrate that the relationship between cost and capability is non-linear at the task level: cheaper models at higher effort levels can catch up to or even surpass more expensive models.

On SWE-bench Pro, Opus 5.5 improves by approximately 2.5 points when moving from medium to high effort, with costs increasing by about 43% (rising from 70% of the high-effort baseline to 100%). Moving from low to medium effort yields an improvement of about 5.5 points, at roughly double the cost of low effort (rising from one-third of the high-effort baseline to 70%). This implies that if you are currently bottlenecked at medium effort, trying high effort is often more cost-effective than jumping directly to Fable 5.2. You should only consider upgrading to Fable 5.2 when Opus 5.5 at high effort remains insufficient.

For scenarios requiring the mixed use of multiple models, the blog offers a core principle: place sub-agents on cheaper models, while keeping judgment calls and code editing on the main model. Sub-agents are suitable for tasks like searching files, running tests, and reading logs—errors here are easy to spot, and judgmental decisions are inherently made by the main model. When the model proves insufficient, Osmani suggests executing steps in order: first confirm whether the model can self-verify (via tests, builds, or scripts calling endpoints), then raise the effort level by one step, and finally switch to a larger model starting at a lower effort level. On an internal defect-search benchmark, Opus 5.5 acting as the main model orchestrated up to 25 Sonnet 5.5 workers to process approximately 22 million tokens of source code. This approach cost about one-third less than running Opus 5.5 alone, with scores remaining essentially flat.

Fontes:claude.dev Blog

Pesquisa

South Korean AI Startup Motif Technologies Completes Bridge Funding Round with Planned NVIDIA Participation for Next-Gen LLM 'Motif 4'

Tópico · 韩国Motif AIMaterial
2026-11-09 16:07 GMT+8

South Korean artificial intelligence foundation model startup Motif Technologies announced on November 9 that it has completed a bridge funding round to finance the development of its next-generation large language model, "Motif 4." According to Platum, the round includes participation from existing investors and chip giant NVIDIA, though the specific amount remains undisclosed. A bridge round typically refers to interim financing secured before a formal subsequent major funding stage to sustain operations or advance critical projects.

"Motif 4" is the successor to "Motif 3," released by the company in August 2026, and is scheduled to be published as an open-weight model in the first quarter of 2027. Open weights mean the model parameters will be publicly available, allowing developers to download, use, and modify them. Motif Technologies has set an ambitious goal for "Motif 4" to enter the top five global models.

Previously, "Motif 3" was developed under the Ministry of Science and ICT (MSIT)'s "Independent AI Foundation Model" project in South Korea. According to Artificial Analysis data, "Motif 3" scored 47 points on the composite intelligence index, ranking 10th among all AI models globally and 4th among all open-weight models. Excluding US and Chinese enterprise models, it held the highest rank. The company stated that by adopting proprietary architectures such as Group Differential Latent Attention (GDLA), it completed the development of "Motif 3" using only 768 GPUs over five months.

However, due to time and resource constraints, "Motif 3" did not undergo sufficient post-training steps like reinforcement learning (RL). Post-training is a crucial phase where pre-trained models are refined for practical usability and human alignment. Motif Technologies plans to utilize the newly secured funds to significantly strengthen the post-training of "Motif 4," aiming to enhance its utility and commercial competitiveness.

Im Jung-hwan, CEO of Motif Technologies and a PhD in mathematics from Oxford University, stated: "In the fiercely competitive global model landscape, survival requires differentiation at every stage, from architecture to post-training." He acknowledged that "Motif 3" had regrets regarding insufficient post-training due to resource limitations and promised that "Motif 4" would deliver results combining technical excellence with superior usability and commercial viability.

Fontes:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Pesquisa

Cisco Research: Trust, Not Technology, Is Now the Speed Limit on Enterprise AI

Tópico · 思科AI信任报告Material
2026-11-10 04:00 GMT+8

At Cisco Live APJC in Melbourne on November 9, 2026, Cisco Systems Inc. released a new research report titled "Permission to Scale: The Critical Role of Trust in Successful AI Adoption." Authored by independent researcher Brad Howarth, the study draws on data from 25 Australian organizations.

The core finding is that while the conversation about enterprise artificial intelligence has long focused on capability—which model is smartest or which GPU is fastest—this constraint has shifted. AI can now do far more than most organizations will allow, and the gap is trust.

Carl Solder, CTO of Cisco ANZ, noted that CIOs and CTOs often gravitate toward infrastructure and data, neglecting the culture pillar. He described a customer who invested heavily in technology but ignored culture, resulting in significant pushback from employees and customers against the toolsets.

The report introduces the concept of "calibrated trust," where users know when to rely on AI, when to verify outputs, and when a human must decide. Anna Tosky, Chief People Officer at Elmo Software, highlighted the risk of "AI slop," where employees send unchecked, erroneous AI-generated reports.

To address this, the report recommends a specific practice for IT leaders: stop treating log-ins as a proxy for trust and instead track how often people challenge or override AI outputs. Usage shows people are clicking; override rates show they are thinking. Leadership buy-in is also crucial, exemplified by Flight Center Travel Group's founder using AI daily to give staff permission to experiment.

Although the sample size is small, the findings align with broader industry discussions, suggesting that the biggest barrier to scaling AI is not technical but people-related.

Fontes:SiliconANGLE

Pesquisa

US Power Firm NRG Energy Reports Strong Q3 2026 Results and Initiates 2027 Guidance

Material
Verificado 2026-11-10 09:45 GMT+8

NRG Energy, Inc. (NYSE: NRG), a US provider of electricity, natural gas, and smart home solutions, announced its financial results for the third quarter ended September 30, 2026, on November 9, 2026. The company reported GAAP Net Income of $561 million and basic GAAP EPS of $2.59. Non-GAAP metrics included Adjusted Net Income of $983 million, Adjusted EPS of $4.68, and Adjusted EBITDA of $2,013 million.

The significant year-over-year improvement in profitability was driven by contributions from the portfolio acquired from LS Power, higher realized capacity prices in the East, and favorable weather impacts in Texas. These benefits were partially offset by increased depreciation, amortization, and interest expenses associated with the acquired assets and related financing.

In addition to reaffirming its 2026 guidance, NRG initiated its 2027 financial guidance for the first time. The 2027 Adjusted EPS range is set at $8.05 to $10.05, with Adjusted EBITDA guided between $5.5 billion and $6.0 billion. It is noted that the current 2027 guidance does not reflect the expected impacts of the previously announced 4.9 GW asset sale.

The company also updated its capital allocation plan. Beyond the 2026 plan of $1 billion in share repurchases and approximately $407 million in dividends, NRG announced a 2027 base plan including $1 billion in debt reduction, $1 billion in share repurchases, and an 8% increase in the annual common dividend to $2.05 per share. Proceeds from the upcoming asset sale are expected to support a pro forma $3 billion share repurchase program across 2026 and 2027.

Fontes:sec.gov

Pesquisa

OpenAI Launches $200 ModRetro Handheld That Lets Users Create Game Boy Games with Codex

Material
2026-11-10 09:19 GMT+8

OpenAI's merchandise store, Supply Co., has officially launched the "ModRetro Chromatic Codex Edition," a handheld game console co-developed with retro hardware manufacturer ModRetro. Priced at $200 (approximately 1,344 RMB), the device pays homage to the classic Nintendo Game Boy with its LCD screen, directional pad, and physical cartridge slot, but its primary feature is deep integration with OpenAI's coding assistant, Codex.

According to official specifications, the bundle includes one programmable Codex game cartridge and a dedicated carrying case. Users do not need traditional programming skills; by installing the ModRetro Chromatic plugin in the Codex desktop app and describing their game ideas in natural language—such as character settings and gameplay logic—the AI can automatically generate game projects, assets, and code. The resulting ROM file can be written directly to the included physical cartridge and played offline on the handheld.

This product marks an attempt by OpenAI to extend AI coding capabilities from software development scenarios into physical consumer hardware. Previously, OpenAI released a mechanical keyboard supporting AI status displays, while this co-branded handheld further lowers the barrier to entry for retro game development. Currently, the product is available only in the United States, Canada, Australia, New Zealand, Japan, the United Kingdom, and the European Union, with no direct shipping to mainland China.

Fontes:爱范儿

Pesquisa

LanceDB Engineer Weston Pace: The Core Disagreement on Table Formats Lies Not in Data Files, but in the Coordination Mechanism for Concurrent Writes

Tópico · 表格式并发协调机制Material
2026-11-09 23:03 GMT+8

Once written, files are essentially immutable, whereas data naturally requires modification. Rows need correction, new observations must be recorded, and old records must be deleted. This contradiction is the sole reason for the existence of table formats: since files cannot be modified, changes are expressed by creating new files. Appending a row writes a new file; deleting a row writes a marker file; compacting data writes a batch of new files; even acquiring and releasing table locks result in new files. The cost is that at any given moment, a single table corresponds to multiple files, some of which are not data files but deletion markers or indexes. Consequently, someone must define a set of rules specifying which files are currently valid and what role each plays.

Writing the rules themselves is not difficult. The true difficulty lies in this: when two writers simultaneously wish to submit their respective sets of new files, how does one become the 'current version' while the other fails? Weston Pace refers to this problem as the 'atomic kernel' of table formats and notes that all table formats—regardless of whether they are called timelines, catalogs, or other names—ultimately rely on some mechanism to synchronize concurrent writers. The choice of this mechanism has profound impacts on the performance and complexity of the table format, and it clearly divides the design space into two categories.

The first category is zero-infrastructure atomics. Almost every cloud storage provider offers put-if-not-exists semantics, while traditional file systems generally support rename-if-not-exists. By leveraging these primitives, a version sequence can be built without deploying any additional services: writers attempt to atomically create the metadata file for the next version number; the winner succeeds, and the loser retries. The benefits are obvious—fifty EC2 instances can simultaneously publish data to the same table on S3 without any coordination service. However, the author provides a specific performance ceiling: as of his writing, cloud storage put-if-not-exists operations are limited to approximately ten commits per second. For scenarios involving high-frequency small writes, this becomes a bottleneck.

The second category is centralized coordination. DuckLake uses a Postgres instance as a commit coordinator, synchronizing writers internally via mutexes or CAS (Compare-And-Swap) primitives. Iceberg's REST catalog also falls into this category. This approach potentially allows for higher transaction rates (though the author admits uncertainty on this point) and works across all file systems, including non-mainstream ones that do not support put-if-not-exists or POSIX rename operations. The cost is the need to maintain a long-running service, which increases operational burden and the attack surface for failures.

These two approaches are not a matter of superiority versus inferiority, but rather different applicability scenarios. Tables with read-heavy workloads suffice with zero infrastructure; tables with high-frequency small writes may require centralized coordination; tables with large-batch writes can use either. Lance defaults to zero infrastructure but allows extension to centralized coordination; the standard path for Iceberg is REST-based coordination, but it also provides a zero-infrastructure 'boring catalog'. The author concludes that table formats face numerous subtle technical decisions, each favoring a certain type of workload. The best practice is often to implement multiple paths and use heuristics to automatically select the appropriate one, rather than forcing users to navigate a configuration nightmare.

Fontes:lancedb.com

Pesquisa

Prioritizing AI Use Cases: Classify by Type First, Measure Data Distance Next; Model Cost Barely Affects Priority

Material
Verificado 2026-11-10 09:14 GMT+8

89% of organizations use AI in at least one function, yet only 37% report an impact on enterprise EBIT. Johnny Mai, Director of Generative AI Product Strategy and Go-To-Market at Qwen, attributes this gap not to model capability but to backlog prioritization—proposals often rank based on who presented last rather than value. His proposed solution follows four steps: classify by type, plot on a quadrant, score with weights, and then allocate funding.

Six types—Drafters, Answerers, Extractors, Routers, Analysts, and Operators—cover nearly all enterprise AI requests, and the type itself predicts effort before any code is written. Drafters and Routers typically require only a model and a prompt; Operators demand permission management, system integration, and error-handling protocols. A more critical structural insight is that many high-value Operators are composed of lower-tier types. For instance, a procurement approval agent begins as an Extractor (reading the request), becomes a Router (selecting the approval path), acts as an Analyst (checking against budget), and finally operates as an Operator. Funding the component parts first reduces the overall cost of the agent.

Mai identifies 'data distance' as the hidden variable: how far a use case sits from data that is already clean, accessible, and permissioned. He claims this metric predicts effort more reliably than model choice. In the Impact × Ease 2x2 matrix, the vertical axis represents Impact, and the horizontal axis represents Ease (not Effort, so the top-right quadrant signifies 'better'). Each quadrant dictates a specific funding action: High Impact/High Ease projects ship immediately with Q1 metrics; Low Impact/High Ease items receive self-serve access and usage policies to observe organic adoption; High Impact/Low Ease items receive funding to reduce their data distance, with gate review dates set instead of launch dates; Low Impact/Low Ease items are parked with documented reasons for revisit when data or models change. A common error is treating 'Park it' like 'Ship now' by setting a launch date for a project whose data does not yet exist.

Scoring utilizes six weighted criteria: three measuring Impact (Value, Volume, Measurability) and three measuring Ease (Data Readiness, Integration, Error Tolerance), each scored from 1 to 5. Weights must be established before reviewing candidates, as they represent value decisions rather than technical ones. Mai emphasizes that scoring should occur in a room with sponsors, data owners, and engineers present. Disagreement regarding data readiness is often the most valuable output of such meetings, surpassing the utility of final numerical scores.

A worked example illustrates why model cost rarely affects prioritization: assuming 40,000 invoices monthly, with 3,000 input tokens and 500 output tokens per invoice, using qwen3.8-flash in the Singapore region (priced at $0.15 per million input tokens and $0.47 per million output tokens) results in a monthly token cost of approximately $27.40. However, Mai notes that the true workload lies in data readiness and integration, potentially requiring two quarters to launch. His conclusion is to 'price the data distance, not the model.' In the worked example, a procurement approval agent ranks fifth on the weighted total but second on Impact; reading the quadrant before the total prevents misclassifying a 'Park it' item as a 'Ship now' candidate.

Fontes:Alibaba Cloud Community

Pesquisa

Cloudflare Launches Public Vulnerability Intelligence Service 'Zero Wire' with AI Agent Prompts

Tópico · Zero WireMaterial
2026-11-10 08:00 GMT+8

Network infrastructure and cybersecurity company Cloudflare recently launched a public vulnerability intelligence service called Zero Wire on its Radar platform. The service is accessible even to non-Cloudflare customers. Rather than being a comprehensive database of all CVEs (Common Vulnerabilities and Exposures), Zero Wire curates major vulnerabilities in software providing internet-facing services, including web applications, related execution environments, and libraries. It specifically focuses on issues where proxy server protections can mitigate attack risks.

The intelligence provided includes details on affected systems, evidence of exploitation, and remediation advice, along with links to vendor security advisories and research materials. For some vulnerabilities, Cloudflare also provides Web Application Firewall (WAF) protection information, though it notes these measures only reduce specific attack risks and cannot replace software patching. Additionally, the service displays changes in HTTP request volumes matching specific WAF rules as an indicator of potential exploitation attempts, while clarifying that this data does not prove successful attacks or determine the scale of actual harm.

Beyond traditional intelligence, Zero Wire offers AI agent assessment prompts. Enterprises can use these prompts to instruct AI agents to perform read-only checks on internal systems to identify if affected software is in use, assess patch status, and review relevant protective measures. The agents are expected to list check evidence, pending questions, and follow-up recommendations. Cloudflare states that copying the prompts themselves does not automatically scan or modify enterprise systems, positioning this as a new tool for proactive self-verification.

Fontes:iThome

Pesquisa

Django Co-Founder Simon Willison: After Replacing strtod with Clinger's Fast Path, C Version of SQLite Extension Surpasses Rust

Visão rápida
2026-11-09 06:26 GMT+8

Django co-founder Simon Willison had a specific question while walking his dog: could he write an SQLite extension function that parses strings like "-$1,234.56" or "'12.5%'" into numeric values that can be summed and sorted? He decided to use Claude (Opus 5.5) to simultaneously explore both C and Rust implementations, comparing performance, code complexity, and artifact size.

The initial results were contrary to his expectations. The Rust extensions were larger—the Rusqlite version was 321 KB and the sqlite-loadable-rs version was 298 KB—while the C extension was only 14 KB, yet Rust was actually faster. The reason was specific: the equivalent floating-point parsing implementation in the Rust standard library is about twice as fast as C's strtod, and the naive C implementation relied on this function to handle decimals.

He then issued a second prompt to Claude: write a custom C function to parse text into decimal numbers to catch up to or exceed Rust. Claude adopted Clinger's fast path—a known optimization strategy that skips general floating-point parsing overhead for common number formats. On 2 million rows of data, the optimized C version took 19 ns/row on mixed data (Rust was 41–46 ns), 16 ns/row on pure integers (Rust was 32–36 ns), and 18 ns/row on pure decimals (Rust was 47–55 ns), taking roughly half the time of Rust in each scenario.

All four implementations passed consistency checks from 8 million fuzz tests, confirming that the parsing results were identical.

This experiment demonstrates a verifiable pattern: when a performance bottleneck is pinpointed to a specific library function, language choice itself is not the decisive factor; targeted algorithmic replacement can bridge language-level gaps. AI's role here was not to discover new algorithms, but to implement a known technique under explicit instructions and verify its correctness.

It should be noted that this was a first-party benchmark test covering only a specific scenario for a single function; the author did not claim the results generalize to broader C vs. Rust comparisons. Clinger's fast path itself is a known algorithm proposed in the 1990s, and its correctness has a broad foundation of practical application.

Fontes:Simon Willison’s Weblog

Pesquisa

Django Co-creator Simon Willison Uses Claude Code to Locate a Mysterious Beep with Multi-device Microphones, Achieving ~40cm Accuracy

Tópico · Claude CodeMaterial
2026-11-10 07:57 GMT+8

Something in the storage room was intermittently beeping. The intervals between the beeps were long enough that standing in the room made it difficult for Simon Willison to pinpoint the source, and he did not want to dig through every cardboard box. So he instructed Claude Code to build a system: using all available devices—two iPads, a smartphone, a MacBook Pro, and a Surface Ultra—to listen simultaneously and triangulate the source of the beep. The system ultimately narrowed the location down to within approximately 40 centimeters; upon moving those boxes, he found an old carbon monoxide detector that desperately needed its batteries replaced.

Willison considers the most interesting part of this project to be not the localization accuracy, but what it proves: you can give a coding agent ears. Claude does not have native audio input, but it can run JavaScript and Python code to analyze audio captured by a microphone. Willison had Claude start a Python server, connect to multiple devices via WebSocket, call the browser WebAudio API on each device to request microphone permissions, and then hot-deploy new audio processing code. This allowed him to essentially step away and let Claude query audio logs, adjust detection algorithms, and reconfigure its 'ears' independently.

The technical mechanism is based on the 2007 BeepBeep paper. Each listening device maintains a 60-second audio buffer and runs a detection script that uses Fast Fourier Transform (FFT) and a high-pass filter to identify beeps. When a device reports a suspected beep, the server selects the device with the lowest round-trip latency to play a short chirp sound, then requests that all devices upload audio clips spanning from approximately 0.6 seconds before the beep to 0.8 seconds after the chirp. This chirp acts as a timestamp, allowing the clips from different devices to be precisely aligned so that time-difference-of-arrival calculations can be performed for triangulation.

During development, Willison initially used Claude Code Cloud on his iPhone with Opus 5.5 to generate the problem description and initial code, then cloned the repository to his laptop to continue iterating with the newly released Fable 5.2. A key design feature was that Opus 5.5 wrote an installModule() method during the initial build, allowing the server to push and execute new JavaScript code to any listening device via WebSocket. This meant Willison did not need to physically visit each device to refresh pages or re-authorize microphones; the agent could deploy modifications remotely. He added features such as drag-and-drop positioning, clap testing, timestamp annotations, audio playback, and markers for the mysterious beep through a series of specific prompts.

However, automatic inter-device ranging did not work well, forcing Willison to add a 'rough mode' where users manually dragged device icons to approximate locations, which the system then fine-tuned. This indicates that the method still relies on manual initial calibration in real-world environments. Additionally, the 40-centimeter accuracy was derived from a single specific scenario—a small room, intermittent beeping, and five devices—and is not a reproducible benchmark result. Willison also noted a similar approach presented by Martin Woodward at GitHub Universe, who gave a coding agent eyes by connecting a badge computer via USB-C and pointing a webcam at its screen to verify if updates succeeded. His conclusion is that finding ways to allow coding agents to perceive the surrounding world is a highly engaging endeavor.

Fontes:Simon Willison’s Weblog

Pesquisa
Próxima →
Entrar / Cadastrar
LeituraPesquisaRadarFramework de investimento

The author lists four patterns that should not be funded as originally pitched: Operators lacking underlying Extractors or Routers (effectively funding three disguised projects as one); use cases without baseline metrics (making it impossible to prove impact the following year); projects requiring vendor engineers for continuous operation (Gartner predicts 70% of enterprises will abandon agentic AI built by vendor forward-deployed engineering by 2028); and projects with low data readiness scores (1 or 2) but high enthusiasm. Limitations to note include: this six-type classification and weighting framework is proprietary to the author and not an independently verified industry standard; the article is published by Qwen/Alibaba Cloud, and the worked examples and product recommendations naturally point to its Model Studio platform; while the methodology claims platform neutrality, readers should verify these claims independently.