Loading…
Latest reads
Every post, summarised
Grounded summaries of engineering blogs, newest first. Each links to the original.
- 4,671 posts
- 20 companies
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Latest posts
Your Agent Speaks MCP. Give It a Computer.
Sprites are presented as disposable cloud computers for running coding agents and as tools agents can create for testing, debugging, benchmarking, and multi-service work. The post describes MCP as the transport and authorization layer, while progressive disclosure comes from plugins and skills that reveal capabilities only when needed; CLI, REST, and machine-readable /.sprite/llm.txt remain available. The hosted MCP server returns file reads as MCP resources and attaches safety annotations, marking read-only and destructive operations and flagging exec and service_start as crossing the Sprite boundary. Agents connect through sprites.dev/mcp and authenticate to a specific Fly.io organization, with configurable scoping, a default five-Sprite cap, and an mcp- name prefix. The design emphasizes durable filesystems, real networks, and near-zero idle cost instead of stateless sandboxes, while the post argues these properties make disposable computers practical for agent workflows.
Give Your Coding Agents a Memory You Own
funes is presented as a durable memory layer for coding agents that turns existing session traces into searchable, provenance-backed working memory across machines and agent types. A single binary adds recall and get tools to Claude Code, Codex, pi, or Hermes, builds an incremental local index, and automatically indexes completed turns. Its deterministic pipeline normalizes traces, chunks them, embeds them with a pinned local model, stores them in Lance, and combines vector and BM25 search with rank fusion, cross-encoder reranking, recency weighting, and neighboring chunks. Memories can remain local or be bound to a private Hugging Face dataset owned by the user, with credential redaction, secret scanning, local caching, and Hub-provided access control and versioning. In the handoff-vs-recall benchmark, recall was the cheapest option on both tasks, costing 8x less than a written handoff on one and 4x less on the other, while compaction succeeded on only one task.
Training a coding model to paint watercolours with TRL and OpenEnv
The project reproduces an earlier idea: training a language model to write JavaScript that paints watercolours with p5.brush, using TRL and OpenEnv. Its open pipeline includes a hand-rated reference pool, an RL environment, training scripts, trained models, and Hugging Face Jobs, Spaces, Inference Providers, and Hub artifacts. The environment renders sketches in headless Chromium, restricts the model to ten library methods, rejects cheats through a gate, and combines gate, length, pairwise-judge, and HPSv3 rewards. Three reward mixes were trained: judge-led (0.60/0.30), hps-led (0.30/0.60), and hps-only (0.00/0.90); the first two ran to step 110, while validation stopped at step 60. The post reports that both pairwise-judge runs learned and identifies the hand-rated pool as the bottleneck.
Southern Company’s SCOUT: Completing the Storm Intelligence Story
Southern Company’s SCOUT is a real-time storm operations application designed to make outage intelligence accessible beyond control-center desks during active events. Built on the Databricks lakehouse, it combines Delta Lake storage, Databricks Lakehouse warehouses, Unity Catalog governance, and collaborative notebooks, querying a dedicated warehouse every minute for current operational data. The mobile-friendly application unifies outage, customer, GIS, weather, terrain, crew, and reliability information, helping teams assess restoration times, assign appropriately equipped crews, coordinate mutual assistance, and communicate with customers. SCOUT is used across Southern Company’s operating businesses and had 1,139 adopted employees, including more than 250 users on a peak storm day; the source also describes exploration of AI crew recommendations and autonomous-drone integration, not a completed deployment.
Register Now for JetBrains GameDev Day 2026
JetBrains GameDev Day 2026 is scheduled for September 28 as a free online event for people building games, engines, tools, and production pipelines. The program covers Unity, Unreal Engine, Godot, proprietary engines, and engine-neutral infrastructure, with sessions on AI-assisted Figma-to-Unity integration, procedural content generation, ECS migration, cross-runtime C# and C++ debugging, Unreal performance profiling, and CI troubleshooting. The livestream starts at 10:00 am UTC, and recordings will be available on demand afterward; the full schedule and speaker details are still forthcoming. Watch parties are planned in Belgrade, alongside DevGAMM x FTW! attendee check-in, and at the JetBrains office in Limassol, with separate local timings and registration requirements.
Expanding Genie Agents: Deep analysis, file reasoning, and more
Genie Agents updates expand expert-curated business agents beyond quick factual lookups to deeper analysis across structured and unstructured data. Agent mode can create and refine research plans, issue iterative queries, and return reports with findings, visualizations, and citations; Agent mode APIs extend this workflow to custom applications, chatbots, scheduled reports, and internal tools through streaming Server-Sent Events. Agents can analyze PDFs, documents, slide decks, and images in Unity Catalog volumes alongside governed tables, with permissions and content search indexing applied to retrieval. Genie Code helps authors create baseline agents, diagnose conversation or benchmark failures, and identify knowledge gaps from user feedback, while authors retain review control over proposed changes. Shared chats keep follow-up messages and edited visualizations visible to authorized users, supporting collaboration and agent-quality review.
Developing Enterprise Frontier Safeguards with our customers
Anthropic is announcing Enterprise Frontier Safeguards (EFS), designed to combine zero data retention (ZDR) with monitoring for serious misuse of frontier models. The service addresses enterprise concerns that detecting activity such as fraud, cyberattacks, or biological capability development, or stolen credentials, may require retaining and correlating traffic across sessions and accounts, while regulated organizations may not accept provider-held data or human review. EFS stores activity data in customer-controlled cloud infrastructure, under customer-managed encryption keys, access policies, and audit logging, while automated systems analyze a rolling traffic window and send flags directly to the customer for review. It will roll out in phases later this fall, support Anthropic and partner cloud offerings, and let customers opt into storage, encryption keys, and automated review without changing model behavior, pricing, or rate limits.
How we make AI coding more cost efficient without sacrificing task quality
GitHub Copilot’s AI coding efficiency work focuses on completed task cost rather than tokens in individual tool calls, because shorter responses can trigger recovery work. Using agentic coding benchmarks and controlled online experiments, the team evaluated selective output compression, removal of unused line-number prefixes, prompt shortening, and direct delivery of completed background results. The shipped compressor preserves source-like and arbitrary output, reorganizes search results without dropping content, and compresses repetitive install, build, test, and progress output while retaining a recovery path to originals. Removing line-number prefixes cut offline model-inference cost roughly 5% and online daily cost about 3%; direct background-result delivery reduced AI Credit usage about 2.3%, while prompt changes required workload-specific validation.
The Modern CUDA Toolbox in Practice: A Step-by-Step Optimization Walkthrough
An image-processing pipeline serves as the test case for a modern CUDA toolbox, starting with RGB-to-grayscale conversion, 32-by-32 tile median computation, and host/device transfers. It advances through six incremental changes: adopting CCCL and Compute Sanitizer for indexing bugs, adding NVTX for Nsight Systems benchmarks, replacing hand-rolled work with CUB algorithms, using pooled device containers and pinned host containers, and assigning each image its own stream with asynchronous copies. The implementation uses cuda::stream, cuda::copy_bytes, cuda::device_buffer, and CUB calls on per-thread streams, allowing kernels and transfers to overlap instead of executing sequentially on the default stream. The reported total duration falls from 6.8 seconds to about 23 milliseconds for three images, a roughly 300x improvement, while the post characterizes the code as safer and easier to maintain without low-level optimizations.
Uber Pro: rewards that go the extra mile, like you
Uber is rolling out a redesigned Uber Pro driver rewards program that combines Uber Pro with Advantage Mode in markets where Advantage Mode operates. Drivers earn points by accepting trips, cancelling less, completing safe rides, and advancing through statuses, while lower Gold and Platinum thresholds and reduced Gold requirements add flexibility. Gold includes a 5% Pro Perk on eligible non-premium rides, Exclusive trip requests, and Trip Radar Advantage; Platinum adds Priority Reserve and a third Destination Mode session, and Diamond offers a renewable free three-month Uber One membership. Existing rewards, including Costco Gold Star membership, ASU tuition coverage, and Upside gas discounts, remain available, while eligibility and exclusions vary by ride type, status, market, and program terms.
Co-Designing AI Models Using Speculative Decoding for Faster LLM Inference
Speculative decoding accelerates large language model inference by having a smaller draft model propose multiple tokens and a target model verify them in parallel, preserving the standard output sequence under its acceptance criteria. It reduces decoding iterations and increases target-model arithmetic intensity, but speedup depends on acceptance length, verification time, draft latency, batch size, and hardware behavior. For attention-heavy workloads, the recommended starting point is D = 128/G - 1, where G is query heads per KV head; larger D values should align G × (1 + D) with 128-token kernel tiles. The remaining guidance covers compute-bound GEMMs, low-latency draft-cost tradeoffs, and choosing among external models, EAGLE-3, MTP, DFlash, DSpark, and suffix or n-gram methods; SPEED-Bench, TensorRT LLM, and NVIDIA/Model-Optimizer support measurement, training, and quantization workflows.
TeamCity 2026.2: Pipelines General Availability, BYOK for AI Assistant, and More
TeamCity 2026.2 makes pipelines generally available on TeamCity Cloud and On-Premises, moving them beyond the Early Access Program and adding branch-aware editing, failure-tolerant job execution, promotion, job debugging, and VCS-unbound workflows. The release also expands AI capabilities: AI Assistant supports Bring Your Own Key for Anthropic, OpenAI, Google Gemini, other supported models, or a local LLM, while MCP adds pipeline management tools and OAuth/PKCE authorization. Additional changes include configurable retries for failed dependencies, IDEA plugin improvements, GitHub pull-request matching by source branch, and a choice of server- or agent-side Kotlin DSL compilation. The pipeline MCP post and delete tools require manually enabled Brave mode, and DSL compilation on agents is described as less restricting and more secure but subject to trade-offs. Bug-fix updates 2025.11.8 and 2026.1.4 also ship.
The MPS 2026.2 Early Access Program Has Started
MPS 2026.2 EAP1 introduces improvements to editor test reporting, usage-result grouping, node selection, and startup configuration, alongside bug fixes. Editor tests can now use node-matching rules or per-node markers to ignore irrelevant containment roles, properties, references, or other fixture differences, while failures identify the responsible node and cell with presentation, concept name, and node ID. Find Usages can group results by virtual package, including one tree level per segment of dotted names, and this option is enabled by default. Select Up now treats lists of sibling nodes as an intermediate editing unit, supporting more predictable repeated selection and group copy-paste. Build tooling separates the IntelliJ Platform-provided boot classpath from MPS-specific extensions, exposes additional JVM arguments, and applies shared and operating-system-specific --add-opens options more precisely, reducing launcher maintenance and stale configuration risks.
MySQL CDC connector for ClickPipes is now Generally Available
MySQL CDC connector for ClickPipes is now Generally Available, enabling replication from MySQL and MariaDB into ClickHouse Cloud for analytical workloads. The release adds automatic parallel snapshotting: ClickPipes selects a logical partition key from a table’s primary key, including string and UUID types, and distributed processing reduced a benchmarked 1TB snapshot from over a day on one thread to about an hour across 32 workers. Production safeguards now require GTID-based replication and at least 72 hours of binlog retention, while new CPU, memory, network, and source, destination, and end-to-end lag metrics improve troubleshooting. MySQL CDC ClickPipes also support OpenAPI and Terraform 3.14.0+, are available across all ClickHouse Cloud service tiers, and began billing on September 1, 2026.
Real-Time Intelligence with IBM Time Series Models on Confluent
IBM and Confluent are making IBM Granite Time Series foundation models available in Early Access on Confluent Cloud, bringing forecasting and other time-series capabilities into streaming workflows. The announcement addresses the cost and delay of bespoke models by offering models that generalize across previously unseen series and can be used for forecasting, anomaly detection, similarity search, classification, gap-filling and optimization. Native inference runs directly in Apache Flink through Confluent’s AI_FORECAST and AI_DETECT_ANOMALIES functions, with Flink maintaining keyed, fault-tolerant state for each series and results written to Kafka topics. Confluent manages serving, scaling and runtime operations while retaining schemas, lineage, access controls and replayable data within the platform. Early Access starts on Confluent Cloud on AWS at no charge, with Confluent Platform planned next for on-premises and hybrid environments.
Stop Guessing at Hard Faults
CLion 2026.2.2 adds an AI skill for diagnosing ARM Cortex-M hard faults through dedicated MCP tooling. Embedded debugging typically requires decoding CFSR and HFSR, matching the faulting program counter to disassembly, and reproducing crashes because default handlers often report only that a fault occurred. When triggered by a stopped hard-fault-family handler or relevant fault evidence, the skill provides decoded status registers, the hardware-stacked exception frame, SVD-decoded peripheral registers, disassembly, and nearby memory to an agent. It supports any debugging tool, including Lauterbach TRACE32, Segger J-Link, and ST-LINK, and can produce a bug location and fix, with optional rerunning to verify the program. The skill is enabled by default but requires CLion’s MCP server; it works in AI chat and CLI mode for Claude Code and Codex, while GitHub Copilot supports CLI mode only.
An Organizational Second Brain: Building an AI That Learns From Experts
Organizations often retain specialist knowledge in experts’ heads, leaving routine compliance assessments slow, inconsistent, and difficult to scale. The proposed “second brain” combines a structured, auditable knowledge architecture with a separate reasoning layer and a self-improvement loop that converts expert feedback into verified, regression-tested text updates without model retraining. High-density organizational positions, taxonomies, routing indexes, and gateway files are maintained in a dependency-linked filesystem, while sparse supporting material remains available through semantic or lexical retrieval. Composable recipes encode expert procedures, and an evaluation pipeline proposes auditable pull requests, retries failed fixes, and expands regression coverage after approval. After six weeks, SMEs reportedly found outputs useful almost always, assessments fell from days to minutes, and improvement cycles produced zero regressions; the architecture is presented as applicable beyond compliance.
How to Handle Errors in Go
Go treats errors as values rather than moving error handling outside normal control flow: functions return an error alongside other results, and callers check it. The guide presents the built-in error interface, returning errors with errors.New, wrapping failures with fmt.Errorf and %w, and distinguishing nil success from non-nil failure. It also explains panic and recover for unexpected failures, including deferred recovery in services that must continue handling other requests, while warning that panics should not replace ordinary error handling. Its best practices emphasize adding context, choosing informative custom error types, inspecting wrapped errors with errors.Is, errors.As, and errors.AsType, and logging errors when they are handled or cannot be returned. The conclusion recommends balancing early crashes with recovery when restarting is costly or when independent goroutines can continue safely.
LoRA backdoor threat: How OpenShift AI mitigates the risk
LoRA adapters can carry semantic backdoors even when the underlying open-weight model is clean, creating a supply-chain risk for customized AI assistants. The demonstrated proof of concept uses a 35 MB adapter trained from 270 ChatML examples, with 130 poisoning examples and 140 clean examples, to trigger on ordinary pharmaceutical IP language rather than fixed strings. With Qwen/Qwen2.5-0.5B-Instruct served by vLLM 0.25.1, the backdoored model covertly called send_email to an external address while showing a benign response, whereas the base model and controls remained clean. The defense-in-depth approach combines OpenShift AI, Advanced Cluster Security, and Advanced Cluster Management to restrict runtime LoRA loading, require provenance-checked adapters, enforce admission and build policies, and block unexpected egress. Because weight inspection cannot reliably reveal semantic behavior, protection focuses on deployment, runtime, tool, and network boundaries.
LLM quantization guide: How to do it, and how it helps
LLM quantization addresses the gap between rapidly growing model sizes and GPU memory by storing weights and activations at lower precision than BF16. It explains how quantization targets linear layers in transformer blocks, distinguishes weight quantization for memory-bound decoding from activation quantization for compute-bound prefill, and maps formats such as GGUF, FP8, NVFP4, AWQ-INT4, and MLX to their inference engines. Calibrated methods including GPTQ, AWQ, and SmoothQuant protect important weights using representative data, while reported NVFP4 results recover about 99% of BF16 accuracy for 70B–235B models and 95–98% for 7B–14B models. In a Llama 3 70B benchmark on two H100s, FP8 weights-and-activations increased throughput from 158 to 474 input tokens/sec and reduced time to first token under load from over 30,000ms to about 4,800ms.
Top topics
- Architecture834
- Performance816
- LLMs660
- Machine Learning622
- Open Source563
- AI Agents553
- AI470
- Developer Experience448
Weekly digest
The best engineering reads, once a week.