Desire a fast and straightforward strategy to watch all of your favourite media at dwelling? You don’t should pay a full paycheck to make it occurs. The Flixy TV Stick is a small and compact choice that turns any HDMI-equipped display into your new TV-based streaming portal.
An reasonably priced streaming follow the whole lot you want for a good time
The Flixy TV Stick is a small distant that delivers massive efficiency. It helps crisp 4K Extremely HD decision with HDR for vivid colours and razor-sharp element, so even dated sitcoms look freshly remastered. A quad-core processor retains navigation clean and cuts load instances; no extra frozen menus whilst you await the subsequent episode to start out. Twin-band Wi-Fi locks onto your house community, and the inner antenna is tuned to scale back dropouts, so binging a sequence feels seamless as an alternative of worrying.
It comes with a distant with a matte shell so that you don’t have to fret about getting fingerprints throughout it. Massive, well-spaced buttons make it straightforward to make use of in low gentle, and a built-in microphone permits you to launch apps or search throughout providers with easy voice instructions. You’ll be able to say issues like “play the most recent superhero film” and the stick provides you with some choices out of your subscriptions in seconds. Should you misplace the distant, you may play a particular locator tone that’s triggered from the companion app in your cellphone. That’s nice information when you lose your streaming remotes or different tech on a regular basis within the sofa.
There’s a companion app you need to use to get to exhibits and films. Select what you need to watch and construct a queue, use your streamer as a second display, and mirror images or movies onto the TV for grou watching. There’s even a lodge mode that walks you thru logging in over captive Wi-Fi portals, that means you may maintain your watchlist rolling on the highway with out lugging round a cumbersome field.
Vivo has introduced the X300 and X300 Professional, its newest flagship telephones, and so they look nearer to the corporate’s much more premium Extremely handsets than ever. On paper, the Professional’s digital camera setup seems simply nearly as good because the X200 Extremely’s, particularly because of providing help for a similar add-on telephoto extender lens launched with that cellphone when it launched in April this 12 months, just a few months after the preliminary X200 handsets.
Cameras are clearly Vivo’s give attention to each new X300 telephones, which every characteristic a brand new 200-megapixel HPB imaging sensor designed in collaboration with Samsung. The X300 Professional makes use of that sensor for its 85mm-equivalent telephoto digital camera, which affords related specs to the industry-leading telephoto on the X200 Extremely, aside from a barely slower f/2.67 aperture. The primary digital camera would possibly even higher that of the Extremely, utilizing the Sony LYT-828, the successor to the sensor present in that cellphone, with a sooner aperture.
It’s solely the ultrawide that’s clearly a step down from the X200 Extremely’s, utilizing a smaller sensor that’s unlikely to ship the identical picture high quality, particularly in low mild. All three rear cameras additionally help 4K, 120fps, 10-bit Log video recording. It’s sufficient to make the X300 Professional really feel like an announcement cellphone from Vivo, channeling the acclaim it’s acquired for its Extremely fashions in the direction of the following step down in worth.
The common X300’s cameras are nearly as spectacular because the Professional’s on paper.Picture: Vivo
The common X300 ought to impress too. Right here, that 200-megapixel Samsung sensor is used for the principle digital camera, paired with a quick f/1.68 aperture. It’s joined by a 50-megapixel telephoto, and the identical 50-megapixel ultrawide and selfie lenses because the Professional.
Each telephones help the telephoto extender, a 2.35x lens that attaches to the telephones’ present telephoto lenses by way of a devoted cellphone case and mount. Paired with the X200 Extremely it enabled me to take some fairly extraordinary long-range pictures that no different cellphone could be able to, and I’m solely shocked to see Vivo rolling help out to its different flagships so quickly, slightly than holding it again as an Extremely-only characteristic. It’s nonetheless offered along with a bigger Pictures Package, together with the mounting case and a digital camera grip with a battery pack inbuilt.
The telephoto extender impressed me after I examined it with the X200 Extremely this 12 months.Photograph: Dominic Preston / The Verge
Each telephones are powered by MediaTek’s flagship Dimensity 9500 chipset, a rival to the Snapdragon 8 Elite Gen 5. A couple of different specs are shared too: 90W wired and 40W wi-fi charging, IP68 and 69 sturdiness rankings, and ultrasonic fingerprint sensors.
The most important distinction, past the digital camera specs, is the scale: whereas the Professional is an enormous cellphone, with a 6.78-inch show, the common X300 has a comparatively compact 6.31-inch display screen. Each telephones are slim, coming under 8mm thick, although the X300’s 7.95mm is a hair’s width thinner. The batteries range too, although Vivo has made some attention-grabbing claims right here: it means that because of effectivity optimizations, the 6,510mAh and 6,040mAh silicon-carbon batteries within the two telephones are comparable in longevity to 7,500mAh and seven,000mAh cells in different {hardware} — hopefully we’ll get the possibility to place that declare to the take a look at ourselves.
The X300 (proper) is available in barely extra enjoyable colours than the X300 Professional (left).Picture: Vivo
Each the X300 and X300 Professional can be found to order in China now, operating the new OriginOS 6, Vivo’s tackle Android. The X300 begins from ¥4,399 (round $620), with the Professional at ¥5,299 (round $745). The telephoto lens is offered individually for ¥1,299 (round $180), or ¥1,499 (round $210) for the total pictures package.
The corporate hasn’t confirmed wider launch plans, however its flagship fashions usually launch elsewhere in Asia together with some European markets — and for the primary time it’s confirmed that the brand new OriginOS will too, lastly changing the dreadful Funtouch OS software program Vivo has used exterior of China for years. An X300 Extremely can be more likely to launch in early 2026, although Vivo must pull out all of the stops to search out room to distinguish it from the X300 Professional.
Observe subjects and authors from this story to see extra like this in your personalised homepage feed and to obtain electronic mail updates.
The vivo X300 sequence is now official. It consists of the bottom X300 with a compact design, and the X300 Professional. Each smartphones are powered by the just lately introduced MediaTek Dimensity 9500 SoC and provide a number of upgrades over their predecessors.
vivo X300 Professional
Beginning with the flagship, the vivo X300 Professional will get an upgraded Zeiss triple rear digital camera setup. It gives a 50MP Sony LYT-828 foremost 1/1.28-inch sensor with f/1.57 aperture and gimbal OIS. The cellphone additionally will get a 50MP Samsung JN1 ultrawide unit, and a brand new 200MP 1/1.4-inch Samsung HPB sensor for its telephoto wants. On the entrance, vivo has included one other 50MP Samsung JN1 sensor for selfies and video calls.
The cellphone is powered by the MediaTek Dimensity 9500 SoC, paired with vivo’s V3+ imaging chip, as much as 16 GB RAM, and 512 GB of storage. It packs a 6,510 mAh battery with assist for 90W wired and 40W wi-fi charging.
It sports activities a 6.78-inch 8T LTPO show that gives a 120Hz refresh fee and 1260 x 2800 pixels decision. The show helps HDR10+, Dolby Imaginative and prescient, and options an ultrasonic fingerprint scanner.
The vivo X300 Professional weighs 226g, is 7.99mm thick, and gives an IP68/IP69 ranking. It’s accessible in Black, Brown, Blue, and White shade choices. Different specs embrace, twin stereo audio system, Bluetooth 6.0, USB-C 3.2 Gen 1 port, and Wi-Fi 6. It runs Android 16-based OriginOS 6.
vivo X300
Shifting on to the smaller X300, the cellphone can be powered by the MediaTek Dimensity 9500 SoC, coupled with as much as 16GB of RAM and 512GB of storage. This yr, the bottom mannequin will get a extra compact physique.
It comes with a 6.31-inch 8T LTPO panel with 120Hz refresh fee and 1216 x 2640 pixels decision. You get an ultrasonic fingerprint scanner as effectively. The cellphone packs in a smaller 6,040 mAh battery with 90W wired and 40W wi-fi charging assist.
The vivo X300 has a 200MP major rear digital camera (Samsung HPB sensor) with a 1/1.4-inch sensor dimension and f/1.68 aperture. It additionally will get a 50MP 1/1.95-inch Sony LYT-602 APO telephoto digital camera and a 50MP Samsung JN1 ultrawide digital camera. The JN1 sensor can be used on the entrance for selfies.
As a result of smaller dimension, the X300 weighs 190g and is 7.95mm skinny. It gives an IP68/IP69 ranking and is accessible in Black, Blue, Pink, Purple, and Purple shade choices. The cellphone additionally comes with OriginOS 6 based mostly on Android 16. Different specs embrace Bluetooth 5.4, NFC, dual-band Wi-Fi 6, and twin stereo audio system.
The vivo X300 Professional is priced at CNY 5,299 ($745) for the 12GB/256GB variant and CNY 5,999 ($845) for the 16GB/512GB possibility. In the meantime, the 16GB/1TB mannequin will price CNY 6,699 ($945)
However, the bottom X300 begins at CNY 4,399 ($620) for the 12GB/256GB possibility. The 16GB/256GB mannequin is priced at CNY 4,699 ($660), the 12GB/512GB variant is CNY 4,999 ($700), 16GB/512GB is CNY 5,299 ($745), and the 16GB/1TB possibility is CNY 5,999 ($845).
Salesforce introduced Monday the newest model of its AI agent platform as the corporate appears to lure enterprises to its AI software program in an more and more crowded market.
The shopper relations supervisor large unveiled the brand new platform, branded Agentforce 360, forward of its annual Dreamforce buyer convention that kicks off October 14. This newer model of Agentforce consists of new methods to instruct AI brokers via textual content, a brand new platform to construct and deploy brokers, and new infrastructure for messaging app Slack, amongst others.
A notable facet of Agentforce 360 is its new AI agent prompting instrument, referred to as Agent Script, which will likely be launched in beta in November. Agent Script offers customers the power to program their AI brokers to be extra versatile and higher reply to “if/then” conditions. This permits AI brokers to be programmed to be extra predictable in much less inflexible conditions like buyer questions.
Customers can faucet into “reasoning” fashions, which declare to assume earlier than responding versus responding based mostly on patterns. Anthropic, OpenAI and Google Gemini energy these “reasoning” brokers.
Salesforce additionally introduced it’s releasing a brand new agent constructing instrument, Agentforce Builder, which permits customers to construct, take a look at and deploy AI brokers from a singular spot. This instrument, which will likely be launched in beta in November, consists of Agentforce Vibes, an enterprise-grade app vibe coding instrument that Salesforce introduced earlier this month.
The corporate additionally introduced a broader integration between Agentforce and Slack. Salesforce mentioned its core apps, together with Agenforce Gross sales, IT and HR, amongst others, will floor straight in Slack beginning this month and broaden via the start of 2026.
Slack is piloting a brand new model of its Slackbot chatbot that’s meant to be extra of a customized AI agent that learns about its person and can supply insights and solutions.
Techcrunch occasion
San Francisco | October 27-29, 2025
Salesforce desires Slack to function an enterprise search instrument sooner or later too and plans to launch connectors with platforms like Gmail, Outlook, and Dropbox in early 2026.
This newest replace from Salesforce comes at an fascinating time for the enterprise AI market. Firms proceed to launch AI options geared toward their enterprise clients whereas enterprises battle to see a return on funding for these instruments.
Final week Google introduced Gemini Enterprise, a collection of instruments — lots of which have been already out there — for constructing enterprise-grade AI brokers, that counts Figma, Klarna and Virgin Voyages as early clients, amongst others.
Anthropic additionally began to point out traction for its enterprise product, Claude Enterprise. The corporate introduced it struck a take care of consulting large Deloitte to deliver its Claude chatbot to Deloitte’s 500,000 international staff — its largest enterprise deal but. Anthropic introduced a strategic partnership with IBM the following day.
Salesforce touts that Agentforce has 12,000 clients — considerably increased than any of its rivals, in keeping with its Agentforce press launch. Early pilot clients of its Agentforce 360 upgrades embrace Lennar, Adecco, and Pearson.
That is all regardless of a latest MIT research discovered that 95% of enterprise AI pilots fail earlier than they attain manufacturing as corporations nonetheless battle to justify spending cash on these AI instruments.
Assume your WAF has you lined? Assume once more. This vacation season, unmonitored JavaScript is a vital oversight permitting attackers to steal cost knowledge whereas your WAF and intrusion detection methods see nothing. With the 2025 procuring season weeks away, visibility gaps should shut now.
Get the entire Vacation Season Safety Playbook right here.
Backside Line Up Entrance
The 2024 vacation season noticed main assaults on web site code: the Polyfill.io breach hit 500,000+ web sites, and September’s Cisco Magecart assault focused vacation customers. These assaults exploited third-party code and on-line retailer weaknesses throughout peak procuring, when assaults jumped 690%.
For 2025: What safety steps and monitoring ought to on-line retailers take now to forestall related assaults whereas nonetheless utilizing the third-party instruments they want?
As vacation procuring site visitors will increase, corporations strengthen their servers and networks, however a vital weak spot stays unwatched: the browser atmosphere the place malicious code runs hidden on customers’ gadgets, stealing knowledge and bypassing normal safety.
The Consumer-Aspect Safety Hole
Current trade analysis reveals the regarding scope of this safety hole:
These statistics underscore a basic shift within the risk panorama. As organizations have strengthened server-side defenses by means of WAFs, intrusion detection methods, and endpoint safety, attackers have tailored by focusing on the browser atmosphere the place conventional monitoring instruments fall brief as a result of following:
Restricted Visibility: Server-side monitoring instruments can not observe JavaScript execution inside customers’ browsers. WAFs and community monitoring options miss assaults that function fully within the consumer atmosphere.
Encrypted Site visitors: Fashionable net site visitors is encrypted by way of HTTPS, making it troublesome for community monitoring instruments to examine the content material of information transmissions to third-party domains.
Dynamic Nature: Consumer-side code can modify its conduct primarily based on person actions, time of day, or different elements, making static evaluation inadequate.
Compliance Gaps: Though laws like PCI DSS 4.0.1 focus now extra on consumer aspect danger, there’s nonetheless restricted steering on client-side knowledge safety.
Understanding Consumer-Aspect Assault Vectors
E-skimming (Magecart)
Maybe probably the most infamous client-side risk, Magecart assaults contain injecting malicious JavaScript into e-commerce websites to steal cost card knowledge. The 2018 British Airways breach, which uncovered 380,000 clients’ cost particulars, exemplifies how a single compromised script can bypass sturdy server safety. The assault operated for 2 weeks undetected, harvesting knowledge instantly from the checkout kind earlier than transmitting it to attacker-controlled servers.
Provide Chain Compromises
Fashionable net purposes rely closely on third-party providers, analytics platforms, cost processors, chat widgets, and promoting networks. Every represents a possible entry level. The 2019 Ticketmaster breach occurred when attackers compromised a buyer assist chat device, demonstrating how a single third-party script can expose a whole platform.
Shadow Scripts and Script Sprawl
Many organizations lack full visibility into all JavaScript code executing on their pages. Scripts can dynamically load different scripts, creating a posh net of dependencies that safety groups battle to trace. This “shadow script” phenomenon signifies that unauthorized code could also be working with out specific approval or monitoring.
Session and Cookie Manipulation
Consumer-side assaults can intercept authentication tokens, manipulate session knowledge, or extract delicate info from cookies and native storage. Not like server-side assaults that go away community logs, these operations happen fully inside the person’s browser, making detection difficult with out specialised monitoring.
Actual-World Vacation Season Assaults: Classes from 2024
The 2024 vacation season offered stark examples of the escalating client-side risk. The notorious Polyfill.io provide chain assault, which started in February 2024 and impacted over 100,000 web sites by the vacations, demonstrated how a compromised third-party script might redirect customers to malicious websites. Equally, the Cisco Magecart assault in September 2024 focused vacation customers by way of their merchandise retailer, highlighting how even giant organizations are susceptible to cost knowledge theft throughout peak intervals.
Past these high-profile incidents, the pervasive nature of client-side threats was evident. The compromised Kuwaiti e-commerce website Shrwaa.com hosted malicious JavaScript information all through 2024, infecting different websites undetected and showcasing the “shadow script” downside. The Grelos skimmer variant additional illustrated session and cookie manipulation, deploying pretend cost types on smaller, trusted e-commerce websites simply earlier than Black Friday and Cyber Monday. These incidents underscore the vital want for sturdy client-side safety measures.
The Vacation Season Amplifies Danger
A number of elements make the vacation procuring interval notably susceptible:
Elevated Assault Motivation: Increased transaction volumes create profitable targets, with Cyber Monday 2024 seeing 5.4 trillion every day requests on Cloudflare’s community, with 5% blocked as potential assaults.
Code Freeze Intervals: Many organizations implement growth freezes throughout peak seasons, limiting the flexibility to reply shortly to newly found vulnerabilities.
Third-Occasion Dependencies: Vacation promotions typically require integration with further advertising and marketing instruments, cost choices, and analytics platforms, increasing the assault floor.
Useful resource Constraints: Safety groups could also be stretched skinny, with most organizations scaling again after-hours SOC staffing ranges by as much as 50% throughout holidays and weekends.
Implementing Efficient Consumer-Aspect Safety
1. Deploy Content material Safety Coverage (CSP)
Begin with CSP in report-only mode to achieve visibility into script execution with out breaking performance:
This method supplies fast insights into script conduct whereas permitting time for coverage refinement.
The CSP Lure to Keep away from: When implementing CSP, you will possible encounter damaged performance from legacy scripts. The tempting fast repair is including `’unsafe-inline’` to your coverage, which permits all inline JavaScript to execute. Nevertheless, this single directive utterly undermines your CSP safety, it is the equal of leaving your entrance door unlocked as a result of one key would not work. As an alternative, use nonces (cryptographic tokens) for professional inline scripts: `
2. Implement Subresource Integrity (SRI)
Be certain that third-party scripts have not been tampered with by implementing SRI tags:
3. Conduct Common Script Audits
Keep a complete stock of all third-party scripts, together with:
Goal and enterprise justification
Knowledge entry permissions
Replace and patching procedures
Vendor safety practices
Different options if the service turns into compromised
4. Implement Consumer-Aspect Monitoring
Deploy specialised client-side monitoring instruments, starting from browser-based CSP validators to Net Publicity administration options to industrial Runtime Utility Self-Safety (RASP) options, that may observe JavaScript execution in real-time, detecting:
Surprising knowledge assortment or transmission
DOM manipulation makes an attempt
New or modified scripts
Suspicious community requests
5. Set up Incident Response Procedures
Develop particular playbooks for client-side incidents, together with:
Script isolation and removing procedures
Buyer communication templates
Vendor contact info and escalation paths
Regulatory notification necessities
Implementation Challenges and Options
Whereas the advantages of client-side safety are clear, implementation can current obstacles. Here is tips on how to navigate frequent challenges:
Legacy System Compatibility
Implement CSP progressively, beginning with highest-risk pages
Use CSP reporting to establish problematic scripts earlier than enforcement
Think about deploying a reverse proxy to inject safety headers with out utility modifications
Monitor that SRI checks add minimal overhead (usually underneath 5ms per script)
Observe actual person metrics like web page load time throughout rollout
Vendor Resistance
Embrace safety necessities in vendor contracts upfront
Body necessities as defending each events’ reputations
Keep a vendor danger register monitoring safety posture
Doc uncooperative distributors as highest-risk dependencies
Useful resource Limitations
Think about managed safety providers specializing in client-side safety
Begin with free browser-based instruments and CSP report analyzers
Prioritize automation for script stock, monitoring, and alerts
Dedicate 6-12 hours month-to-month for preliminary setup and ongoing monitoring, or funds 1-2 days quarterly for complete audits in enterprise environments with 50+ third-party scripts
Organizational Purchase-In
Construct enterprise case round breach prices (common Magecart assault: $3.9M) versus monitoring funding ($10K-50K yearly)
Organizations with devoted client-side monitoring detect breaches 5.3 months quicker than trade common (lowering the 7.5-month detection window to 2.2 months), considerably limiting knowledge publicity and regulatory penalties
Current client-side safety as income safety, not IT overhead
Safe government sponsorship earlier than vacation freeze intervals
Emphasize prevention is much less disruptive than responding to an energetic breach throughout peak season
Trying Ahead
Consumer-side safety represents a basic shift in how we method net utility safety. Because the assault floor continues to evolve, organizations should adapt their safety methods to incorporate complete monitoring and safety of the consumer atmosphere.
The vacation procuring season supplies each urgency and alternative: urgency to handle these vulnerabilities earlier than peak site visitors arrives, and alternative to implement monitoring that may present precious insights into regular versus suspicious script conduct.
Success requires shifting past the standard perimeter-focused safety mannequin to embrace a extra complete method that protects knowledge wherever it travels, together with inside the person’s browser. The organizations that make this transition is not going to solely defend their clients through the vacation rush however set up a extra resilient safety posture for the yr forward.
Discovered this text fascinating? This text is a contributed piece from one in all our valued companions. Observe us on Google Information, Twitter and LinkedIn to learn extra unique content material we publish.
Every time it involves coaching mannequin, firms often guess of feeding it increasingly more knowledge for coaching.
Greater datasets = smarter fashions
When DeepSeek launched initially, it challenged this strategy and set new definitions for mannequin coaching. And after that got here a wave of mannequin coaching with much less knowledge and optimized strategy. I got here throughout one such analysis paper: LIMI: Much less Is Extra for Clever Company and it actually bought me hooked. It discusses the way you don’t want hundreds of examples to construct a strong AI. In truth, simply 78 fastidiously chosen coaching samples are sufficient to outperform fashions educated on 10,000.
How? By specializing in high quality over amount. As an alternative of flooding the mannequin with repetitive or shallow examples, LIMI makes use of wealthy, real-world situations from software program growth and scientific analysis. Every pattern captures the total arc of problem-solving: planning, software use, debugging, and collaboration.
The outcome? A mannequin that doesn’t simply “know” issues: it does issues. And it does them higher, sooner, and with far much less knowledge.
This text explains how LIMI works!
Key Takeaways
Company is outlined because the capability of AI programs to behave autonomously, fixing issues by means of self-directed interplay with instruments and environments.
The LIMI strategy makes use of solely 78 high-quality, strategically designed coaching samples centered on collaborative software program growth and scientific analysis.
On the AgencyBench analysis suite, LIMI achieves 73.5% efficiency, far surpassing main fashions like GLM-4.5 (45.1%), Kimi-K2 (24.1%), and DeepSeek-V3.1 (11.9%).
LIMI reveals a 53.7% enchancment over fashions educated on 10,000 samples, utilizing 128 instances much less knowledge.
The research introduces the Company Effectivity Precept: machine autonomy emerges not from knowledge quantity however from the strategic curation of high-quality agentic demonstrations.
Outcomes generalize throughout coding, software use, and scientific reasoning benchmarks, confirming that the “much less is extra” paradigm applies broadly to agentic AI.
What’s Company?
The paper defines Company as an emergent functionality the place AI programs perform as autonomous brokers. These brokers don’t look forward to step-by-step directions. As an alternative, they:
Actively uncover issues
Formulate hypotheses
Execute multi-step options
Work together with environments and instruments
This contrasts sharply with conventional language fashions that generate responses however can’t act. Actual-world purposes like debugging code, managing analysis workflows, or working microservices, require this type of proactive intelligence.
The shift from “considering AI” to “working AI” is pushed by trade wants. Firms now search programs that may full duties end-to-end, not simply reply questions.
Why Much less Information Can Be Extra Efficient?
For over a decade, AI progress has adopted one rule: scale up. Greater fashions. Extra tokens. Bigger datasets. And it labored: for language understanding. Nonetheless, current work in different domains suggests in any other case:
LIMO (2025) demonstrated that complicated mathematical reasoning improves by 45.8% utilizing solely 817 curated samples.
LIMA (2023) confirmed that mannequin alignment could be achieved with simply 1,000 high-quality examples.
However company is totally different. You possibly can’t study to construct by studying tens of millions of code snippets. You study by doing. And doing properly requires dense, high-fidelity examples: not simply quantity.
Consider it like studying to cook dinner. Watching 10,000 cooking movies would possibly train you vocabulary. However one hands-on session with a chef, the place you chop, season, style, and modify, teaches you tips on how to cook dinner.
LIMI applies this concept to AI coaching. As an alternative of accumulating limitless logs of software calls, it curates 78 full “cooking periods,” each a whole, profitable collaboration between human and AI on a posh process.
The outcome? The mannequin learns the essence of company: tips on how to plan, adapt, and ship.
The LIMI Method: Three Core Improvements
LIMI’s success rests on three methodological pillars:
Agentic Question Synthesis
Queries aren’t generic prompts. They simulate actual collaborative duties in software program growth (“vibe coding”) and scientific analysis. The group collected:
60 real-world queries from skilled builders and researchers.
18 artificial queries generated from GitHub Pull Requests utilizing GPT-5, making certain authenticity and technical depth.
Trajectory Assortment Protocol
For every question, the group recorded full interplay trajectories, multi-turn sequences that embody:
Mannequin reasoning steps
Device calls (e.g., file edits, API requests)
Environmental suggestions (e.g., error messages, person clarifications) These trajectories common 42,400 tokens, with some exceeding 150,000 tokens, capturing the total complexity of collaborative problem-solving.
Deal with Excessive-Affect Domains
All 78 coaching samples come from two domains that symbolize the majority of information work:
Vibe Coding: Collaborative software program growth with iterative debugging, testing, and gear use.
Analysis Workflows: Literature search, knowledge evaluation, experiment design, and report technology.
This focus ensures that each coaching instance is dense with agentic alerts.
Dataset Building: From GitHub to Human-AI Collaboration
The LIMI dataset was constructed by means of a meticulous pipeline:
Step 1: Question Pool Creation
Actual queries got here from precise developer and researcher workflows. Artificial queries have been derived from 100 high-star GitHub repositories, filtered for significant code modifications (excluding documentation-only PRs).
Step 2: High quality Management
4 PhD-level annotators reviewed all queries for semantic alignment with actual duties. Solely the very best 78 have been chosen.
Step 3: Trajectory Technology
Utilizing the SII CLI surroundings, a tool-rich interface supporting code execution, file system entry, and net search: human annotators collaborated with GPT-5 to finish every process. Each profitable trajectory was logged in full.
The result’s a compact however extraordinarily wealthy dataset the place every pattern encapsulates hours of sensible problem-solving.
Analysis: AgencyBench and Extra
To check LIMI’s capabilities, the group used AgencyBench, a brand new benchmark with 10 complicated, real-world duties:
Vibe Coding Duties (4):
C++ chat system with login, associates, concurrency
Java to-do app with search and multi-user sync
Net-based Gomoku recreation with AI opponents
Self-repairing microservice pipeline
Analysis Workflow Duties (6):
Evaluating LLM efficiency on DynToM dataset
Statistical evaluation of reasoning vs. direct fashions
Dataset discovery on Hugging Face
Scientific perform becoming to excessive precision
Information-rich variants: Fashions educated on CC-Bench (260 samples), AFM-WebAgent (7,610), and AFM-CodeAgent (10,000)
On AgencyBench, LIMI scored 73.5%, far forward of all rivals:
First-Flip Practical Completeness: 71.7% vs. 37.8% (GLM-4.5)
Success Price (inside 3 rounds): 74.6% vs. 47.4%
Effectivity (unused rounds): 74.2% vs. 50.0%
Much more putting: LIMI outperformed the ten,000-sample mannequin by 53.7% absolute factors, utilizing 128 instances fewer samples.
On generalization benchmarks, LIMI averaged 57.2%, beating all baselines and data-rich variants. It achieved prime scores on coding (92.1% on HumanEval) and aggressive outcomes on software use (45.6% on TAU2-retail).
Case Research: Actual-World Efficiency
The paper consists of detailed case comparisons:
Gomoku Recreation (Process 3): GLM-4.5 failed at board rendering, win detection, and AI logic. LIMI accomplished all subtasks with minimal intervention.
Dataset Discovery (Process 7): GLM-4.5 retrieved much less related datasets. LIMI’s selections higher matched question necessities (e.g., philosophy of AI consciousness, Danish hate speech classification).
Scientific Operate Becoming (Process 8): GLM-4.5 reached loss = 1.14e-6 after a number of prompts. LIMI achieved 5.95e-7 on the primary attempt.
NBA Reasoning (Process 9): GLM-4.5 usually failed or required most prompts. LIMI solved most subtasks with zero or one trace, utilizing fewer tokens and fewer time.
These examples illustrate LIMI’s superior reasoning, software use, and flexibility.
Machine autonomy emerges not from knowledge abundance however from strategic curation of high-quality agentic demonstrations.
This challenges the trade’s reliance on large knowledge pipelines. As an alternative, it means that:
Understanding the essence of company is extra vital than scaling knowledge
Small, expert-designed datasets can yield state-of-the-art efficiency
Sustainable AI growth is feasible with out monumental compute or knowledge prices
For practitioners, this implies investing in process design, human-AI collaboration protocols, and trajectory high quality: not simply knowledge quantity.
The LIMI paper delivers a daring message: you don’t want 10,000 examples to show an AI tips on how to work. You want 78 actually good ones. By specializing in high-quality, real-world collaborations, LIMI achieves state-of-the-art agentic efficiency with a fraction of the info. It proves that company isn’t about scale. It’s about sign.
As AI strikes from chatbots to coworkers, this perception will probably be essential. The longer term belongs to not those that acquire essentially the most knowledge, however to those that design essentially the most significant studying experiences.
Within the age of agentic AI, much less isn’t simply extra. It’s higher!
Hey, I’m Nitika, a tech-savvy Content material Creator and Marketer. Creativity and studying new issues come naturally to me. I’ve experience in creating result-driven content material methods. I’m properly versed in website positioning Administration, Key phrase Operations, Net Content material Writing, Communication, Content material Technique, Enhancing, and Writing.
Login to proceed studying and luxuriate in expert-curated content material.
A vector database, which shops doc embeddings, scales rapidly and helps distributed storage for superior indexing and vector querying.
A vector library, which is a quicker, lighter method to maintain vector embeddings.
Vector help built-in into the prevailing database to retailer vector embeddings and help querying.
Your best option relies on your particular circumstances. For instance, a vector-native database is probably the most strong technique, but it surely’s too costly and resource-heavy to be sensible for smaller organizations. A vector library is quicker and greatest for instances when latency is the enemy, whereas integrating vector capabilities is best however doesn’t scale effectively sufficient for heavy enterprise wants.
3. Construct a stable retrieval course of.
It’s proper there within the identify – RAG is all about retrieving the best information to construct correct responses. Nonetheless, you may’t merely level your RAG infrastructure at information sources and anticipate it to retrieve the most effective solutions. You want to educate RAG methods tips on how to retrieve related data, with a robust emphasis on relevance. Too typically, RAG methods over-collect information, leading to extreme noise and confusion.
“Experimental analysis confirmed that retrieval high quality issues considerably greater than amount, with RAG methods that retrieve fewer however extra related paperwork outperforming most often people who attempt to retrieve as a lot context as potential, leading to an overabundance of knowledge, a lot of which could not be sufficiently related,” observes Iván Palomares Carrascosa, a deep studying and LLM venture advisor.
Remodeling how LSEG clients join with monetary information in Microsoft options
Enabling brokers inbuilt Microsoft Copilot Studio and deployed in Microsoft 365 Copilot with LSEG- licensed information
Advancing safe, seamless use of LSEG-licensed monetary information in Microsoft options utilizing the Mannequin Context Protocol (MCP) open-source customary
LONDON and REDMOND, Wash.— Oct. 13, 2025 — LSEG and Microsoft Corp. on Monday introduced the following step of their multiyear, strategic partnership: remodeling entry to LSEG-licensed monetary information and powering the deployment of agentic AI with information that’s safe and accessible by way of an LSEG-managed MCP server. By this collaboration, brokers inbuilt Microsoft Copilot Studio and deployed in Microsoft 365 Copilot could be enabled with LSEG information.
By connecting LSEG’s trusted content material with superior AI capabilities, monetary providers professionals can profit from clever instruments that improve decision-making and productiveness and decrease obstacles to innovation — empowering organizations to make use of AI responsibly and securely.
At the moment’s announcement builds on LSEG’s AI technique — LSEG In every single place — which is delivering trusted licensed information to scale AI in monetary providers. The unparalleled depth, breadth and high quality of LSEG’s AI Prepared Content material and taxonomies embrace datasets stretching again over a long time and whole greater than 33 petabytes. Producing insights from these datasets allows Microsoft 365 Copilot customers to make assured selections quicker.
Powering the construct, deployment and integration of customized AI brokers instantly into workflows
Utilizing a brand new MCP server, LSEG clients will be capable to join with information, licensed by way of LSEG merchandise like Workspace and Monetary Analytics, to construct AI brokers in Copilot Studio. Copilot Studio gives a safe, low-code platform for customers to create, customise and deploy superior AI brokers that may be built-in instantly into workflows by way of Microsoft 365 Copilot and different purposes. Copilot Studio allows makers to construct subtle brokers with ease, utilizing frontier AI fashions in a totally managed SaaS expertise that helps a variety of enterprise connectors. Copilot Studio empowers organizations with sturdy governance controls and is natively built-in into Microsoft 365 Copilot, enabling safe and compliant customization at scale.
By Copilot Studio, customers will be capable to entry the monetary information and analytics licensed by way of LSEG instantly inside their on a regular basis workflows to compose subtle AI brokers, combining insurance policies, prompts, instruments and actions. This integration unifies human experience and huge language mannequin (LLM) reasoning with LSEG’s unmatched monetary information, content material and analytics in a safe, ruled, low-code surroundings. LSEG and Microsoft are presently working with clients to construct their first brokers in Copilot Studio powered by LSEG information.
MCP server enabling safe, seamless connectivity and industrywide interoperability
LSEG clients may also reap the benefits of MCP for enhanced connectivity and a unified information and AI ecosystem. LSEG has carried out this superior protocol to make its information accessible and optimized for integration into AI enabled workflows, enhancing how LLMs interpret and use LSEG’s information. An MCP server allows the connection of Copilot Studio with LSEG APIs in addition to full interoperability with clients’ personal AI methods and different third-party purposes.
This functionality allows the switch of datasets in a standardized surroundings, decreasing costly and time-intensive integration of a number of applied sciences for purchasers. This radically simplifies and hastens the supply of AI-enabled instruments and purposes which might be safe, ruled and scalable. LSEG information might be accessible by way of MCP in a phased rollout, beginning with LSEG Monetary Analytics.
The distinctive mixture of Microsoft’s cloud and AI platform with LSEG’s monetary information and analytics delivers differentiated outcomes for trade professionals by way of deeper insights and extra well timed, knowledgeable decision-making in a quickly evolving market.
David Schwimmer, Group Chief Government Officer, LSEG
“LSEG’s partnership with Microsoft is remodeling entry to information for monetary professionals with innovative, AI-driven innovation at scale. LSEG clients can construct, deploy and scale agentic AI instantly into their workflows with safe, seamless connectivity by way of MCP.”
Nick Parker, Microsoft Chief Enterprise Officer and President, Worldwide Gross sales & Options
“Our continued partnership with LSEG underscores a shared dedication to redefining the way forward for monetary providers by way of safe, AI-driven innovation. By combining LSEG’s trusted market information with Microsoft’s cloud and AI capabilities, we’re empowering clients to unlock deeper insights, speed up decision-making and streamline advanced workflows.”
About LSEG
LSEG is a number one international monetary markets infrastructure and information supplier, enjoying an important social and financial function on the earth’s monetary system. With our open strategy, trusted experience and international scale, we allow the sustainable progress and stability of our clients and their communities. We’re devoted companions with intensive expertise, deep data and a worldwide presence in information and analytics; indices; capital formation; and commerce execution, clearing and threat administration throughout a number of asset courses. LSEG is headquartered in the UK, with vital operations in 65 international locations throughout EMEA, North America, Latin America and Asia Pacific. We make use of over 26,000 individuals globally, greater than half positioned in Asia Pacific. LSEG’s ticker image is LSEG.
About Microsoft
Microsoft (Nasdaq “MSFT” @microsoft) creates platforms and instruments powered by AI to ship revolutionary options that meet the evolving wants of our clients. The expertise firm is dedicated to creating AI accessible broadly and doing so responsibly, with a mission to empower each particular person and each group on the planet to realize extra.
DroneShield’s DroneSentry resolution has been recognised by the UK Authorities’s Nationwide Protecting Safety Authority (NPSA).
NPSA gives UK Authorities company normal analysis for counter-drone methods, offering confidence to potential consumers of the answer following NPSA’s rigorous testing.
DroneShield counter-drone options can be found throughout the UK solely by means of BT
DroneShield (ASX:DRO) is happy to announce its DroneSentry multi-layered counter-drone resolution has been efficiently evaluated by the Nationwide Protecting Safety Authority (NPSA). NPSA is an company of the UK Authorities and the DroneSentry resolution is now included inside their Catalogue of Safety Tools (CSE).
DroneShield’s intensive vary of counter-drone options can be found solely by means of BT throughout the UK.
Oleg Vornik, DroneShield CEO, commented, “That is an thrilling step ahead and a public endorsement of the standard and efficiency of our options, for each the UK and different markets. It comes at a vital geopolitical time of quickly escalating drone incidents throughout Europe, as our clients search to deploy counter-drone safety throughout a variety of services, ranging from a really low or no current deployments because of the nascent nature of the sector.”
Dave Pankhurst, Future Related Options Director, BT commented, “BT has labored with DroneShield for a lot of years within the UK, coupling our deep understanding of buyer wants with DroneShield’s know-how. This NPSA recognition highlights DroneShield’s dedication to growing vanguard counter-drone options with essentially the most strong safety credentials.”
DroneShield’s DroneSentry resolution is underpinned by an AI-enabled command-and-control engine, DroneSentry-C2, which fashioned a part of the NPSA analysis and recognition. The C2 engine, a SaaS product, allows integration of a variety of sensors and effectors, to optimise for the tip person necessities and funds.
The DroneSentry resolution includes of an upfront buy of {hardware}, mixed with one or a number of SaaS packages. DroneSentry options have been deployed thus far in Europe, Asia Pacific, the U.S., South America and Australia.
About DroneShield DroneShield (ASX:DRO) is an Australian publicly listed firm with headquarters in Sydney, Australia and Warrenton, Virginia. DroneShield gives counter-drone (C-UAS) safety, specializing in Radio Frequency (RF) sensing, Synthetic Intelligence (AI) and Machine Studying (ML), Sensor Fusion, Digital Warfare (EW), Speedy Prototyping, and MIL-SPEC manufacturing. DroneShield’s capabilities are used to guard navy, authorities, legislation enforcement, vital infrastructure, industrial and VIP safety, all through the world. Be taught extra at droneshield.com.
Associated
Uncover extra from sUAS Information
Subscribe to get the most recent posts despatched to your electronic mail.
The ABB Robotics product portfolio consists of the OmniCore controller. Supply: ABB
One of many world’s largest industrial automation suppliers is altering house owners. ABB Group at the moment mentioned it’s divesting its ABB Robotics & Discrete Automation division to SoftBank Group Corp. for $5.375 billion.
The Zurich-based firm had introduced plans to spin off the unit throughout an earnings name in April. On the time, it famous that orders and revenues had fallen sharply from 2023 to early 2025.
With 2024 revenues of $2.3 billion, ABB Group mentioned ABB Robotics represented about 7% of its revenues and had an operational EBITDA (earnings earlier than curiosity, taxes, depreciation, and amortization) margin of 12.1%. The division has about 7,000 workers, whereas ABB mentioned it has a complete of about 110,000 worldwide.
ABB Robotics’ U.S. headquarters and manufacturing facility are in Auburn Hills, Mich. In March, it mentioned it deliberate to take a position $120 million in increasing manufacturing within the U.S. The corporate’s portfolio consists of industrial automation, force- and power-limited robotic arms, and autonomous cellular robots (AMRs) acquired with ASTI Cell Robotics for $191 million in 2021.
“SoftBank’s provide has been rigorously evaluated by the board and government committee and in contrast with our authentic intention for a spin-off,” acknowledged Peter Voser, chairman of ABB. “It displays the long-term strengths of the division, and the divestment will create instant worth to ABB shareholders.”
“ABB will use the proceeds from the transaction according to its well-established capital allocation ideas,” he added. “Our ambitions for ABB are unchanged, and we’ll proceed to give attention to our long-term technique, constructing on our main positions in electrification and automation.”
SoftBank beefs up robotics portfolio — once more
“SoftBank’s subsequent frontier is bodily AI,” asserted Masayoshi Son, chairman and CEO of SoftBank Group. “Along with ABB Robotics, we’ll unite world-class know-how and expertise below our shared imaginative and prescient to fuse synthetic tremendous intelligence and robotics —driving a groundbreaking evolution that can propel humanity ahead.”
Not solely has SoftBank invested in robotics corporations, but it surely has additionally owned a number of through the years, with combined outcomes. The Tokyo-based firm‘s robotics unit additionally produces the Whiz ground cleaner.
SoftBank acquired humanoid service robotic maker Aldebaran Robotics for $100 million in 2012, rebranded it as SoftBank Robotics in 2016, and bought it as SoftBank Robotics Europe to United Robotics Group in 2022. In June 2025, Aldebaran went into receivership.
In 2018, SoftBank invested $375 million in Zume Pizza, which shut down two years later. In 2019, it gave Nuro.ai $940 million, and the driverless supply automobile maker reached a $6 billion valuation in August 2025.
SoftBank acquired legged robotic developer Boston Dynamics from Alphabet in 2017 and bought it to Hyundai Motor Group for $880 million in 2021.
As well as, SoftBank took a 40% stake in warehouse automation supplier AutoStore for $2.2 billion in 2021. In 2023, SoftBank merged with Berkshire Gray in a money deal price about $375 million.
“SoftBank will likely be a superb new dwelling for the enterprise and its workers,” mentioned Morten Wierod, ABB CEO. “ABB Robotics will profit from the mixture of its main know-how and deep trade experience with SoftBank’s state-of-the-art capabilities in AI, robotics and next-generation computing.”
Editor’s be aware:RoboBusiness 2025, which will likely be subsequent week in Santa Clara, Calif., will function tracks on bodily AI, design and growth, enabling applied sciences, enterprise, area robotics, and humanoids.
Because of the sale of its robotics division, ABB plans modify its reporting construction and transfer to 3 enterprise areas within the fourth quarter of 2025. As introduced in April, the corporate will transfer its Machine Automation division from the robotics group into its Course of Automation one.
ABB Group mentioned the divestment of the robotics division will end in a non-operational pre-tax e-book achieve of roughly $2.4 billion with anticipated money proceeds, web of transaction prices, of about $5.3 billion. It expects separation prices of round $200 million, about half of which was already included in its 2025 steering. ABB estimated transaction-related money tax outflows for native enterprise carve-out of $400 million to $500 million.
The businesses mentioned they count on the transaction to shut, topic to regulatory approvals and customary closing situations, in mid-to-late 2026. The Robotic Reporthas reached out to ABB for additional remark, and we’ll replace this text when and if we hear again.
President Sami Atiya to go away
ABB additionally introduced that Sami Atiya, president of its Robotics & Discrete Automation enterprise space and a member of its government committee, will go away ABB by the top of 2026. He’ll step down from ABB’s government committee on the finish of 2025 and can proceed to help the robotics enterprise and the carve-out course of in 2026 as a strategic advisor.
“It has been an honor to serve ABB, and I wish to thank our workers in addition to clients for the shut partnership since I joined ABB,” Atiya mentioned. “I’m satisfied that with the acquisition by SoftBank, our robotics enterprise will likely be in wonderful palms, making a platform for development that can profit from the quite a few alternatives in automation and AI that we’re seeing throughout many industries and geographies.”
Wierod added: “Since becoming a member of ABB nearly 10 years in the past, Sami has performed an instrumental function in constructing a robotics enterprise that covers the whole lot from industrial robots to state-of-the-art collaborative and autonomous cellular robots, backed by a number of the most superior software program and AI options.”
Sami Atiya is stepping down as president of ABB Robotics & Discrete Automation. Supply: ABB Group
