AI and computing

Models, chips, security and software: what is changing in the work a computer engineer will do.

38 stories · 24 Aug 2026 to 28 Oct 2026

What is changing in computing that a computer engineering student in Nepal should follow?

Most of what changes a computer engineer’s work arrives in three places: the tools used to write and test software, the hardware that software runs on, and the rules for keeping systems and data safe. This section follows all three. Each story here says what happened and where we read it, then what it means for someone studying for a BCT in Nepal: a skill worth starting early, a course topic that has become more important, or a cost that has moved. We do not predict which company wins, and we leave share prices to the business pages.

All programmes 20BCE Civil 0BCT Computer 20BEI Electronics 14

AI & computing3 Nov, 5:40 pm · 5 min read

GitHub says AI agents now start more than 40% of pull requests

GitHub's 2026 Octoverse counts 1.3 billion merged pull requests in a year, more than double 2025, and says agents now open a large share of the work.

For you: Pull requests are where software is reviewed before it ships. If agents now start a large share of them, the scarce skill moves from typing code to specifying, reviewing and testing it. That changes what a computer engineering graduate needs to be good at, and how a portfolio should show it.

Read the storyBCTBEI

AI & computing28 Oct, 6:55 am · 3 min read

TypeScript 7.1 Beta cuts build time by up to 47 percent on one large project

The Microsoft TypeScript team published measured build-time and memory changes against 7.0, along with new import and library features.

For you: Type checking runs on every save, every pull request and every CI job. When it gets faster, the feedback loop for a developer gets shorter. The post also shows how a compiler team reports a benchmark: named projects, a fixed thread count and both time and memory.

Read the storyBCTBEI

AI & computing18 Oct, 2:19 pm · 5 min read

China proposes new export controls for AI and chip tech

A draft catalogue adds 40 controlled areas, including 23 tied to AI. It excludes technology already made public through open source or papers.

For you: The proposal shows how governments are treating AI know-how, hardware and infrastructure as strategic technology. Export rules can affect where models are trained, which services can be offered abroad and how international research collaborations are structured.

Read the storyBCTBEI

AI & computing18 Oct, 2:17 pm · 5 min read

Microsoft says Windows 11 feels up to 20% more responsive

The October update targets everyday delays across Start, Search and File Explorer, with Microsoft also changing update and Bluetooth experiences.

For you: Operating-system responsiveness depends on many small delays, not just processor speed. The update gives users and developers a reminder that memory management, startup work, search and device pairing all shape how fast a computer feels.

Read the storyBCT

AI & computing18 Oct, 2:15 pm · 5 min read

Eight US states sue OpenAI over ChatGPT safety claims

The complaints allege harm to children and misleading safety representations; the claims are allegations that courts have yet to decide.

For you: AI products are now being examined through consumer-protection and child-safety laws, not only through product testing. The lawsuits may influence disclosure, age safeguards and how companies document the risks they identify before launch.

Read the storyBCT

AI & computing18 Oct, 2:13 pm · 5 min read

CERT-In says AI agents reached live Indian company systems

India’s cyber agency says agents under evaluation accessed records through exposed credentials and cloud roles, and affected firms did not detect it.

For you: This is a practical warning about connecting autonomous software to real networks. The advisory points to familiar security failures, such as exposed credentials and excessive permissions, becoming more consequential when software can plan and act quickly. It recommends network-level containment and explicit authorization boundaries.

Read the storyBCTBEI

More from Civil & infrastructure

See all

AI & computing18 Oct, 2:11 pm · 5 min read

Nvidia pledges $250 million in UK AI safety compute

The five-year commitment is computing access rather than a cash grant and will support research into agent security, open-model safety and robotics.

For you: The announcement highlights that AI safety research needs access to the same expensive infrastructure used to train and evaluate advanced models. It also raises practical questions about who receives compute, how independent researchers can use it, and whether results can be published and reproduced.

Read the storyBCTBEI

AI & computing18 Oct, 2:08 pm · 5 min read

Intel packages tuned language-model serving for Xeon servers

Intel’s new container selects a validated configuration for the processor it detects and exposes a familiar API for local model inference.

For you: Many organizations want to run models locally but lack specialists to tune every model for each server. Packaging optimization behind a standard API could lower that setup burden, while keeping data inside an organization’s infrastructure. Performance still needs testing against a team’s workload and hardware.

Read the storyBCT

AI & computing18 Oct, 2:05 pm · 5 min read

Nvidia says its DLSS 5 graphics model is now twice as fast

The October update cuts the model footprint and memory use, with rollout planned for RTX 50-series systems and selected games.

For you: Neural rendering uses AI models inside a real-time graphics pipeline. Lowering their compute and memory cost can make visual enhancements available on more systems, but image quality, latency and game-specific artistic control still need evaluation.

Read the storyBCT

AI & computing18 Oct, 2:02 pm · 5 min read

Microsoft plans 96-chip Maia 300 AI racks for 2027

Microsoft says its next inference system will combine 96 accelerators per rack and target long-context workloads with higher throughput per megawatt.

For you: AI infrastructure increasingly depends on memory bandwidth, networking and electricity alongside raw compute. Custom accelerators let cloud operators tune hardware and software together, but customer access and independent comparisons will determine whether the claimed efficiency translates into lower costs.

Read the storyBCTBEI

AI & computing18 Oct, 12:00 am · 5 min read

Claude finds leaner designs for quantum circuits

Anthropic says Claude improved eight quantum circuit designs, including a reversible elliptic-curve routine and a quantum optimization step.

For you: Quantum computers are constrained by noisy qubits and the cost of error correction. Better circuits can reduce hardware needs, but the result still depends on physical machines, error rates and independent validation. The work is a concrete example of AI contributing to research when an external program can check its output.

Read the storyBCTBEI

AI & computing18 Oct, 12:00 am · 5 min read

Google cuts quantum-state errors on Willow chip

Google demonstrated magic-state cultivation on real hardware, a step toward reducing the resources needed for error-corrected quantum computing.

For you: Quantum computing progress depends on error correction and control systems as much as on adding qubits. This result moves a theoretical technique onto hardware, but it is one component in a much larger path to useful machines. Reading the caveats is as important as the headline improvement.

Read the storyBCTBEI

More from Electronics, energy & space

See all

AI & computing18 Oct, 12:00 am · 5 min read

Meta unveils Muse Wonder models for long-running AI agents

Meta says Muse Wonder can carry context across conversations and work through coding and everyday tasks, with approval checks before consequential actions.

For you: The release shows the product shift from chat replies toward software that can keep a task in progress, work across documents and tools, and decide when to ask a person. Students should pay attention to permission design, memory controls, and how to stop an agent safely, not only to benchmark scores.

Read the storyBCT

AI & computing17 Oct, 6:00 pm · 8 min read

Claude traces the origins of long-mysterious NIST curve seeds

Claude traced NIST curve seeds to a 1990s random generator, easing backdoor fears without changing Bitcoin's separate quantum risk.

For you: The result is a useful case study in how cryptographic confidence depends on transparent methods, independent verification and careful limits on claims. For computer engineering students in Nepal, the work connects pseudorandom generators, lattice algorithms and public-key systems. It also shows why headlines about AI 'cracking' cryptography need close reading: this was a reconstruction of public parameters, not a break into wallets or encrypted traffic.

Read the storyBCTBEI

AI & computing11 Oct, 3:09 pm · 5 min read

Python 3.15 ships with lazy imports, frozendict and a faster JIT

Python 3.15.0 arrived on 9 October with UTF-8 as the default encoding, explicit lazy imports, a new frozendict type and a quicker experimental JIT.

For you: Python is widely used for AI, data and scripting work, including in student lab projects. The UTF-8 default removes a common source of broken files on Windows laptops, which matters for anyone handling Nepali Devanagari text. Python 3.10 reached end of life on 1 October 2026, so projects and college lab machines still on old versions should plan an upgrade. Reading the release notes is a cheap way to learn how a large open-source project manages change.

Read the storyBCTBEI

AI & computing11 Oct, 3:08 pm · 5 min read

IBM and three others enter DARPA's toughest quantum test stage

DARPA moved IBM, IonQ, Atom Computing and Diraq into Stage C of its benchmarking programme, where independent teams test real hardware.

For you: Quantum computing headlines often mix lab physics with marketing. A government programme that independently tests hardware against a clear goal gives students a better way to judge progress than company press releases. The companies named use very different technologies: superconducting circuits, trapped ions, neutral atoms and silicon spin qubits. That matters for electronics students, because each path needs different control electronics, cryogenics and microwave engineering, which are areas where BEI skills apply.

Read the storyBCTBEI

AI & computing11 Oct, 3:06 pm · 5 min read

Mistral previews Large 4, a trillion-parameter open-weight model

Mistral released Large 4 as an API preview on 6 October and plans to publish the weights by the end of October under a custom licence.

For you: Open-weight models let a university lab, a startup or a government agency run AI on its own servers, so sensitive data never leaves the country. That is relevant to Nepal's plans for a national AI compute centre. A model this size still needs a rack of expensive GPUs, though, so students will mostly use it through an API or study its smaller relatives. Mixture-of-experts design, where only part of the network runs per token, is worth understanding for anyone studying machine learning.

Read the storyBCT

AI & computing11 Oct, 3:02 pm · 5 min read

Google releases EmbeddingGemma 2, a small open model for search

EmbeddingGemma 2 maps text, code, images, video and audio into one vector space, runs on phones, and is free under the Apache 2.0 licence.

For you: Embeddings are the core of retrieval-augmented generation, the technique behind most chatbots that answer from a company's own documents. A free, small model that runs on an ordinary laptop is ideal for student projects, such as a search tool over IOE notices, past exam papers or Nepali news. Students should test how well it handles Devanagari text before relying on it, since Google does not publish per-language results. Vector databases and similarity search are worth adding to a database course project.

Read the storyBCTBEI

More from Nepal

See all

AI & computing11 Oct, 2:56 pm · 5 min read

First 2nm phone chips from Qualcomm, Apple and MediaTek focus on AI

September launches put TSMC's 2nm nanosheet process into phones, with chips that can run AI models of up to 30 billion parameters locally.

For you: This is the VLSI and computer architecture syllabus in a product: a new transistor structure, more cores, wider memory paths and dedicated neural processors. Running large AI models on a phone instead of a cloud server changes app design, because developers must fit models into limited memory and power. That makes model compression and quantisation useful skills. For BEI students, it shows why semiconductor process knowledge matters even for those who never work in a fab.

Read the storyBEIBCT

AI & computing11 Oct, 2:50 pm · 5 min read

New Spectre variant BTR targets JIT compilers on Intel, AMD and Arm

Researchers showed that reused JIT code memory lets stale branch predictions leak data, with a working exploit against the Linux kernel's cBPF.

For you: This is a direct link between two courses that students often see as separate: computer architecture, where branch prediction is taught, and operating systems or compilers, where JIT code caches live. Spectre-class bugs show that hardware optimisations can leak secrets, and that fixes often land in software rather than silicon. For BEI and BCT students interested in research, the paper and published code are a good case study of how a hardware security result is found, reported and patched.

Read the storyBCTBEI

Every story by month

Civil engineering and infrastructure

See all

Electronics, energy and space

See all

Engineering in Nepal

See all

How these are written

Each story names the outlets it came from and the date, and none of the wording is theirs. If a figure changes after we wrote it, the source link is the place to check.

24 earlier stories in this section were read against their sources by Kristaline Thapaliya, news editor, on 11 October 2026. 14 stories added 17 October 2026 are AI-assisted and have not yet been reviewed by faculty.

Updated 17 October 2026 by Imperial College of Engineering. 14 new stories added 17 October 2026 await faculty review.