State of coding agents 🧑‍💻, running LLMs locally 💻, just learn SQL 🎓

As of mid-2026, the terminal has become a favored platform for coding agents, with 35 actively maintained CLI agents available ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌  ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ ‌ 

Sign Up |Advertise|View Online TLDR  TLDR Dev 2026-07-06 🧑‍💻 Articles & Tutorials Hunting a 16-year-old SQLite bug with TLA+: is dqlite affected? (14 minute read) SQLite recently addressed a long-standing bug related to the Write Ahead Log (WAL) that could cause database corruption, present in the system for 16 years. The dqlite team investigated whether this bug affects their implementation by modeling SQLite's behavior in TLA+ and comparing it with how dqlite interacts with SQLite and concluded that dqlite is not affected due to its stricter locking mechanisms that prevent simultaneous writes and checkpoints. Agentic test processes, LLM benchmarks, and other notes on agentic coding from Galapagos Island (92 minute read) The exploration of coding agents and their testing capabilities shows a disconnect between the efficiency of AI in automating processes and the reliability of its outputs, especially in coding and debugging tasks. While AI models can speed up workflows and increase code quality through good testing methods, they often struggle with generating reliable tests and identifying the true source of bugs. PostgreSQL and the OOM Killer: Why We Use Strict Memory Overcommit (10 minute read) Strict memory overcommit in PostgreSQL is necessary for preventing catastrophic out-of-memory (OOM) kills, as it allows the database to handle memory allocation failures gracefully without impacting all connections. A kernel bug in Ubicloud's product temporarily disrupted this configuration, showing the importance of careful management of memory limits to guarantee system stability while running PostgreSQL in production environments. 🧠 Opinions & Advice Better Models: Worse Tools (12 minute read) Recent iterations of Anthropic's models, specifically Opus 4.8 and Sonnet 5, show a troubling trend of producing malformed tool calls when interacting with Pi's editing tool, as they frequently append extraneous fields that do not conform to the expected schema. This regression seems to be linked to the models' training environments, which appear to favor a more forgiving approach to tool calls. Write code, not specs (3 minute read) Good code should be self-documenting, allowing for an understanding of its functionality without extensive specifications. Instead of maintaining both code and specifications, focusing on writing code that establishes patterns and relying on tests to validate functionality allows for greater trust and efficiency in development. Some new agentic patterns (10 minute read) At Prime Radiant, new "agentic" patterns have been implemented to enhance collaboration between agents and human implementers, resulting in improved tools and workflows. Notable examples include an agent named Nora, which acts as a team member rather than just an assistant, and a framework allowing multiple agents to collaboratively address tasks, leading to efficient problem-solving and feature development. 🚀 Launches & Tools Plannotator (GitHub Repo) Plannotator is a browser-based tool that enables users to annotate and review plans, specifications, and code changes from various AI coding agents. It integrates with multiple platforms, supports real-time collaboration, and allows for easy sharing of annotated content with team members. Bramble (GitHub Repo) Bramble is a local-first, open-source password manager that encrypts and stores user data on personal devices without relying on cloud servers, allowing for secure peer-to-peer syncing and various authentication methods, including biometric options and passkeys. 🎁 Miscellaneous State of CLI Coding Agents, Mid-2026 (29 minute read) As of mid-2026, the terminal has become a favored platform for coding agents, with 35 actively maintained CLI agents available. The main players like Claude Code, Codex CLI, and Omp have come up as leading options, each having features such as memory capabilities, editing precision, and orchestration. Performance per dollar is getting faster and cheaper (6 minute read) The release of GLM-5.2 shows advancements in inference performance, with 2626 tok/s/node on AMD MI355X at more than twice the cost-efficiency compared to similar NVIDIA setups. Despite historical challenges with AMD's software support, improvements in model optimization are narrowing the gap. ⚡ Quick Links jamesob's guide to running SOTA LLMs locally (10 minute read) The guide shows detailed instructions and recommendations for setting up a state-of-the-art local machine learning system using specific hardware configurations to run advanced LLMs properly. What ORMs have taught me: just learn SQL (8 minute read) ORMs can provide convenience in working with SQL but often complicate data retrieval and management due to issues like attribute creep, inefficient queries, and dual schemas. Rebuilding Cognition's Agentic MapReduce (5 minute read) Cognition's approach to scaling agent tasks, using Agentic MapReduce, enables parallel processing of security findings by sharding codebases, verifying vulnerabilities, and using Git for communication and fault tolerance. If you're a button, you have one job (4 minute read) The differences in how iPhones and Nothing Phones handle image rotation shows the importance of designing user interfaces that accommodate situational disabilities and improve usability by preventing unnecessary wait times for animations. Love TLDR? Tell your friends and get rewards! Share your referral link below with friends to get free TLDR swag! https://refer.tldr.tech/d0af4569/3 Track your referrals here. Want to advertise in TLDR? 📰 If your company is interested in reaching an audience of web developers and engineering decision makers, you may want to advertise with us. Want to work at TLDR? 💼 Apply here, create your own role or send a friend's resume to jobs@tldr.tech and get $1k if we hire them! TLDR is one of Inc.'s Best Bootstrapped businesses of 2025. If you have any comments or feedback, just respond to this email! Thanks for reading, Priyam Mohanty, Jenny Xu & Ceora Ford Manage your subscriptions to our other newsletters on tech, startups, and programming. Or if TLDR Dev isn't for you, please unsubscribe.