Everyone is a Dev. Find the best AI dev tools. Build in public. Join the community. Newsletter → everydev.ai/newsletter
California, USA
- Tweets 3,674
- Following 217
- Followers 381
- Likes 432
Pinned Tweet
As more of the coding workflow can be handed to agents, the hard part is deciding what they can check, and what still needs human eyes before it ships.
Our digest for Oct 18–23: Replit's Monitoring Agent attempting production repairs, Mintlify Autopilot opening docs PRs, VS Code comparing Copilot / Claude / Codex attempts side by side, and Microsoft Scope turning tasks into repeatable agent experiments.
Delegation and approval are separate decisions.
Full rundown:
everydev.ai/p/news-weekly-ai…
1
121
Juggling separate gateways for your APIs and your LLM calls? Apache APISIX from @TheASF handles both. It's an open-source gateway built on NGINX and etcd, with hot-reloadable plugins, load balancing and circuit breaking. Its AI proxy adds fallback and load balancing across LLM providers, plus token-based rate limiting. It works with OpenAI, Anthropic, AWS Bedrock, DeepSeek and Ollama.
1
20
Agents that write to a shared knowledge store can corrupt it, and rolling back is painful. Omnigraph from @modernrelay is an open-source graph database on object storage with git-style branching and row-level three-way merge, so agents work on a branch. It fuses vector, full-text and graph search with rrf(), and runs on S3, R2, GCS or MinIO in open Lance format.
2
1
2
48
Rewriting your code every time you switch LLM providers, or juggling a key per vendor? Antimatter from @tryantimatter is one gateway with a single API key for hundreds of models, compatible with OpenAI chat, Anthropic messages and Responses formats. It adds routing by cost or speed, fallback chains and eval-based model picks. Pay-as-you-go at upstream price + 10%, and it works with Claude Code and Cursor.
1
23
Vector search getting slow or memory-hungry once your index outgrows RAM? USearch from @unum_cloud is an open-source, single-file similarity search engine with HNSW and exact search. It quantizes down to i8 or even b1, serves big indexes from disk via memory mapping, and uses uint40 IDs to handle 4B+ entries. It's already used in ClickHouse, DuckDB, ScyllaDB, TiDB, YugabyteDB and Memgraph.
1
22
Release notes, PR replies and postmortems that read like a chatbot wrote them? sepia is an Agent Skill from @Nyanako0129 that removes AI tells by repairing narrative structure first, then discourse flow, then surface style. It offers write, review, refactor and recreate operations, per-model fingerprints for Claude, GPT, Gemini, DeepSeek and Kimi, and venue-matched rules. MIT licensed, and it works with Claude Code, Codex, Cursor, Cline, Windsurf, Copilot, OpenCode and goose.
1
24
Stuck gluing LLM APIs into a Java codebase with hand-rolled HTTP clients? @langchain4j is an open-source library for building LLM apps on the JVM. It gives one API across 20+ model providers and 30+ embedding stores, plus chat memory, tool calling with MCP support, agents and RAG. Apache-2.0, with Quarkus, Spring Boot, Helidon and Micronaut integrations.
2
1
45
New on EveryDev.ai: llm-d
Open-source, Kubernetes-native distributed LLM inference serving stack that orchestrates vLLM and SGLang across.
1
30
Intelligent prefix-cache and load-aware routing + Kubernetes, pricing, and setup details: @_llm_d_ everydev.ai/tools/llm-d
12
Waiting on GPU quota or paying for idle H100s by the hour? Hyperstack from @Hyperstackcloud rents on-demand NVIDIA GPU VMs with per-minute billing, plus reserved capacity, spot VMs and on-demand Kubernetes. It's paid, with H200 SXM, H100 SXM, A100 SXM and A4000 options, and a Terraform provider with Go and Python SDKs.
1
36
Changing a prompt or swapping a model and just hoping nothing broke? OpenAI Evals gives you a way to test it. It's an open-source framework and registry of benchmarks for LLMs and apps built on them. Write custom or private evals from your own data using JSON and YAML, including model-graded ones. Built by @OpenAI, with Weights & Biases support and optional Snowflake logging.
2
1
125
Need a model to make small calls in your pipeline, like yes/no checks, multiple-choice picks or ratings, without prompting a general LLM and parsing messy text? Jeeves from @posthog is a 9B reasoning model that thinks first, then returns a decision with calibrated probabilities. It's open source, runs on CUDA or Apple Silicon, and ships a Jev-compatible API built on Qwen3.5-9B.
1
1
99