Context Engineering

OpenAI released GPT-6 Sol and GPT-6 Luna at API prices roughly half those of their GPT-5.6 predecessors. The models add controllable prompt-cache breakpoints and are rolling out in Codex and ChatGPT Work.

Teknium’s public evaluation says a Jev compaction strategy removes tool calls and eventually stops yielding savings. Repeated compaction can invalidate caches and increase total token costs, according to the critique.

Tencent released Hy4 Preview, a 770B-parameter mixture-of-experts model with 49B active parameters and a 1M-token context window. vLLM added day-zero support, while Cline, OpenCode Go, and Vercel AI Gateway made the model available.

The Prefix Sliding paper introduces an inference method that preserves the task prefix and a recent-token window while discarding older reasoning tokens. Its authors report up to 3× faster inference without retraining and longer reinforcement-learning rollouts.

Glean says its runtime routes enterprise-agent work across more than 40 models using company context. It reports $0.58 per query and 78% user preference over Claude Cowork in a 180-person benchmark.

A study of 557 coding-agent sessions finds instruction files and working notes account for most of what agents read. Related work puts installed skills' standing prompt cost at 50–280 tokens, while Backpass turns past sessions into reviewable AGENTS.M files.

A University of Pennsylvania study found common compactors retained only 17% of standing session rules while preserving task information. Practitioners use verified Markdown handoffs and background compaction to manage long coding-agent runs.

Codex now lets ChatGPT subscribers enable a 1 million-token context window for GPT-5.6 Sol. Automatic compaction starts at 900,000 tokens, and tokens beyond the default window count double against limits.

A practitioner report found speculative-decoding acceptance fell from 0.71 to 0.18 beyond about 32K context. Separate DSpark and mlx-dspark tests reported speedups on RTX and Apple Silicon setups.

The Echo Gap paper found self-improving agents can store wrongly self-scored episodes. Tested models endorsed 31% to 54% of their own wrong answers, while other work proposed RL-trained harness state and in-model memory.

Practitioner threads argued coding agents work better with task-specific harnesses, compact runbooks, evals, and deterministic checks. Examples included Sentry MCP traces and state-machine guardrails.

Practitioner posts argued agent evals should check final world state and tool-call trajectories, not just single outputs. A 288-run AGENTS.md test found context files did not improve correctness.

A paper summary found agent skills caused regressions across nearly 6,000 paired office-automation runs. VS Code also added prompt-to-skill migration, making the workflow more accessible despite reliability cautions.

A UW study found coding agents can refuse malicious memory-file instructions while preserving the payload for later use. The state-management debate also covered callable memories, reusable skills, and session-history eval pipelines.

Anthropic said Claude Code cut more than 80% of its system prompt without hurting coding evals. Other workflows use filesystem research folders, session IDs, wiki memory, and composable memory APIs to keep context durable.

Practitioners are sharing smaller, staged coding-agent loops instead of one large prompt. Examples include a Claude Code software factory, project-specific review agents, searchable logs, and harness layers for evals and observability.

Firecrawl introduced a new /search system that returns relevant excerpts instead of full pages for agent workflows. The launch claims 94.7% on SimpleQA and roughly 10x fewer tokens than full-page processing.

Developers described agents writing learned context into linked Markdown pages instead of repeatedly rebuilding it from RAG chunks. Related posts argued Markdown docs and llms.txt can improve agent navigation.

Baseten reports that injected facts can retain log-probability lift in weights while disappearing from answers, with prompts recovering 77–80% accuracy. OpenWiki adopted Open Knowledge Format for portable memory files.

Practitioners reported coding agents following outdated agents.md files and generic project instructions. Other posts tied long-context drift and weak domain language to cleanup-heavy coding runs.

Matt Pocock argued teams should delete docs that restate code and keep ADRs, glossaries, and thin navigation pointers. Zeeg reported colocated repo specs drifted and did not steer agents enough.

Practitioners described GitHub or folder-based markdown knowledge bases that feed persistent company or personal context to Codex, Claude Code, and Hermes. OpenWiki added codebase and personal brain modes for the same pattern.

One Claude Code user parsed 95 sessions and estimated subagent prompt caching made total spend about 8% high. pxpipe separately rendered dense text as images to cut context cost, with exactness tradeoffs.

Sakana listed 11 ICML 2026 papers covering LLM speed, memory, and agent evaluation. The lineup includes TwELL sparse kernels, RePo context positioning, CoffeeBench long-horizon agents, SoftMatcha 2 corpus search, Doc-to-LoRA memory, and Fast-weight Product Key Memory.

Condense.chat opened a compression proxy that strips tokens with Helene 1 and compacts settled agent loops with Adeline 1 to about 9% of their size. The service claims 100M saved tokens and 3× plan extension for Claude or Codex users, so test it on non-sensitive workflows first.

A day after /goal and thread automations landed in Codex, practitioners started standardizing on /goal specs, /fork or /side detours, and /rewind plus /compact recovery. The pattern matters because verifier design and compaction timing now control how well long runs hold together.

Junior’s first memory system cut one analytics task from about 3 minutes to 1 minute in early tests, with tokens down two-thirds and tool calls down 60%. The feature moves persistent task learning into the agent loop, though the results are still internal.

Builders released reusable loop artifacts this week, including a Loop Library Skill, repo templates, and published control-loop definitions for docs sweeps, onboarding checks, and error triage. It matters because teams are turning one-shot prompting into persistent agent runs with explicit stop conditions and shared repo state.

Plannotator 0.21.0 added direct document editing, an embedded CLI agent for co-context, first-class HTML handling, and Bedrock and Vertex support. It matters because the tool is moving from annotation-only review toward persistent IDE-like agent sessions with shared file diffs and cloud model backends.

Perplexity rolled out Brain, a self-updating context graph that carries prior sessions, files, and decisions into new Computer tasks. In research preview for Max users, it matters because Perplexity says the memory layer improves correctness and recall while lowering per-task cost on history-dependent work.

Two days after Factory 2.0 launched, Factory added AutoWiki, which turns a repo into structured engineering docs that update on every push and can be generated from Droid with /wiki. The feature keeps architecture and request-path context aligned with the live codebase.

Z.ai released GLM-5.2 MIT-licensed open weights with 1M context and broad runtime support. Vendor and arena results put it near frontier closed models on long-horizon coding.

Codex users are having the agent write its own `/goal` and sub-agent goals, with OpenAI-side commentary describing that as a built-in meta-prompting pattern. The workflow turns long autonomous runs into a tighter control loop, but users still review goals first so a bad objective does not burn tokens for hours.

Users are using Fable 5 as a planner and long-run orchestrator while pushing implementation and heavy reasoning to Opus and Codex. The setup keeps Fable on supervision and planning, so teams can track execution through live status pages on larger tasks.

Hyperbrowser shipped a Claude Code harness, InsForge showed a Fable run drop from 5.5M to 2.3M tokens, and Higgsfield published new MCP workflows. These tools add reusable harness, context, and interface layers around Fable for more controlled runs.

Matt Pocock's /teach skill installs with npx skills add mattpocock/skills --skill teach and runs structured strategy lessons inside a Claude agent. Follow-up posts add primary-source reading to the lessons and point to a larger dedicated repo.

Anthropic published Fable-specific guidance for Claude Code and API, emphasizing the /model switch, higher default effort, simpler prompts, and /goal-style verification loops. The Claude Code team says older prompt scaffolds can work against Fable's longer sessions.

Practitioners shared repeatable setups for multi-hour Claude runs using auto approvals, dynamic workflows, cloud sessions, and critique loops. One large-codebase sweep reported 144 bugs fixed in about four hours with fewer false positives under model critique.

Codex usage moved further into phone-first workflows, with iOS dictation loops, background voice capture, and app updates like searchable settings and restored state. The comparisons still flag rough spots in multi-thread UX, Windows support, and cases where CLI tabs or cloud agents are easier to manage.

A community workflow broke long-running Codex goals into six required fields, then added an eight-item preflight checklist and helper tools. The structure is meant to reduce runs that drift, stop early, or claim completion without an objective verification step.

OpenAI rolled out a more capable ChatGPT memory system that keeps context across conversations, shows a reviewable memory summary, and doubles memory for US Plus and Pro users. The change matters because persistent context becomes a first-class product feature with explicit controls instead of a static saved-memories note list.

Weaviate introduced Engram, a dedicated agent memory service with async writes, semantic topic grouping, tenant scopes, and composable pipelines. It matters because teams can add a hosted memory layer for agent stacks without stitching custom memory workflows into each application.

GitHub Copilot now supports a 1M-token context window and configurable reasoning levels in VS Code, Copilot CLI, and the Copilot app. It matters because larger repo context and explicit effort controls arrive in the same surfaces teams already use for agent-assisted development.

Builders added /dynamic orchestration, custom-model routing, and repo runbooks around Codex as users exposed new session lifecycle controls in the app. That makes Codex a better fit for long-running, multi-context coding work.

Opus 4.8 can accept new system-role instructions after a user turn while keeping earlier prompt segments cacheable. That lets long-running agents update constraints mid-loop without replaying the full system prompt on every call.

A new MeMo paper and several community memory systems converged on keeping knowledge outside the base model through recipe files, semantic and autobiographical stores, and background reconsolidation. The pattern matters because engineers are treating context loss as a systems problem instead of only asking for larger context windows.

Independent developers released browser-control MCP tooling, repo-context graphing and packaging utilities, and token-compression helpers for coding agents. The cluster matters because agent workflows are now adding browser control, context packing, and cost controls as external infrastructure instead of waiting on raw model upgrades alone.

Practitioners published reusable Codex workflows for project audits, memory-driven skill packaging, mobile delegation, and remote computer use. Try the prompt-and-steps patterns if you want to adapt Codex across repos and devices.

Independent builders published reusable skills infrastructure across coding agents, including Project Think preview support, handoff docs, and an htmx v4 skill pack. That matters because skills are starting to work like portable workflow units instead of one-off prompt snippets inside a single tool.

Independent Codex users published Obsidian memory setups, reusable skill prompts, auto-triage flows, and Cloudflare-backed runners for longer jobs. That matters because Codex is being wrapped into persistent workspaces and operator-defined subagents instead of one-shot chats.