Executive Summary
The AI developer ecosystem saw major momentum this period, highlighted by Nous Research's Hermes Agent reaching 100,000 GitHub stars in just 53 days, making it the #1 trending repository. Key integrations are accelerating, with Ollama adding GitHub Copilot CLI support and Hermes Agent integrating X API CLI skills for Twitter interaction. Local AI inference continues advancing, with new quantization techniques enabling Qwen3.5-9B to run 262k context on 8GB GPUs. Meanwhile, Vercel disclosed a security incident involving unauthorized internal system access, and Hugging Face announced expanded agent capabilities with access to 1M+ Spaces.
Key Events
- Hermes Agent hits 100K stars – Nous Research's open-source AI agent milestone, reaching #1 trending repository status in 53 days → tweet
- Vercel security incident – Unauthorized access to internal systems; users advised to review secret management practices → tweet
- Ollama adds GitHub Copilot CLI – Local inference platform now supports terminal-based GitHub operations including PR/issue exploration and code navigation → tweet
- X API CLI skill lands in Hermes Agent –
/xurl <prompt>command enables Twitter actions and search directly within the agent → tweet - TurboQuant llama.cpp fork enables 262k context on RTX 3070 – Benchmarking shows significant VRAM efficiency gains for local inference → tweet
- Hugging Face expands agent platform – Now supporting 1M+ Spaces for agents to leverage → tweet
- CodexBar 0.21 released – Includes Abacus AI provider, Codex Pro $100 support, and performance fixes → tweet
Analysis
Rising Tide in Agent Frameworks: The explosive growth of Hermes Agent (100K stars) signals a shift toward unified agent platforms that combine coding, research, and content capabilities. The rapid addition of skills (infographics, X API, GitHub integration) demonstrates a plug-in ecosystem model gaining traction.
Local Inference Maturation: TurboQuant and advanced quantization techniques (UD-IQ3_XXS) are making 8GB consumer GPUs viable for long-context local models. This trend toward edge AI deployment continues to accelerate, with M4 Max 128GB setups enabling Qwen3.6-35b locally.
Security Concerns Escalating: Vercel's security incident highlights ongoing risks in developer infrastructure. The community response emphasizing Doppler and secret management tools reflects growing awareness of supply-chain security in AI workflows.
What to Watch: Continued convergence of agent frameworks with development tools (Copilot CLI, codex-cli); potential impact of Apple M5 Ultra on local LLM hardware landscape; Hermes Agent's trajectory as it crosses 100K stars.
Tweet Feed
AI Models & Research
@Teknium · 2026-04-19T05:42
Hermes Agent just hit 100,000 stars on GitHub!!! Thank you everyone!! https://t.co/fCqkTXDkDS → tweet link
@Teknium · 2026-04-19T05:55
RT @NousResearch: 100,000 Stars, #1 Trending Repo this Month This is just the beginning. Stay tuned. https://t.co/HLR0yIjiS2 → tweet link
@sudoingX · 2026-04-19T05:59
holy shit. within a month nous research went from 4.5k gitHub stars to 100k and completely ate the lobster. now climbing past 100k and going to eat the rest as the top general agent from science to content to philosophy to code, all baked into one hermes agent. → tweet link
@Teknium · 2026-04-19T04:46
RT @tonysimons_: Just shipped Hermes Vault v0.1.0. It's a local-first credential vault, scanner, and broker for @NousResearch Hermes agent → tweet link
@louszbd · 2026-04-19T04:19
RT @migtissera: Have you tried GLM-5.1? Was using it this morning thinking that my coding harness was set to Opus-4.7. → tweet link
@sudoingX · 2026-04-19T05:43
fuck woke up to a wild morning anon. opened my hermes agent fork. upstream landed in the exact area i've been building for weeks. 1,444 commits behind. 11 conflicts. few hours of resolution, rebase clean. → tweet link
@Teknium · 2026-04-19T02:33
Seems the twitter cli skill is working well ha → tweet link
@Teknium · 2026-04-19T12:27
RT @ResearchWang: https://t.co/KQqwgy2XhL → tweet link
@Teknium · 2026-04-19T12:39
RT @Jaaneek: Use X API through Hermes agent → tweet link
@Teknium · 2026-04-19T12:01
RT @0x_kaize: Hermes Agent hit 100,000 stars on GitHub It took just 53 days! → tweet link
@Teknium · 2026-04-19T17:48
RT @boxmining: 🤖 Not all AI models are built equal for Hermes Agent. After 1 month of testing MIMO V2 Pro, GPT 5.4, Kimi 2.5, DeepSeek 3.2 → tweet link
@Teknium · 2026-04-19T17:39
RT @ToolUsePodcast: I talked to a group of people who build on Hermes Agent and asked them why they chose it over everything else. → tweet link
@gospaceport · 2026-04-18T22:22
RT @NousResearch: Honored to announce we are partnering with Jim Liu to port over his wildly popular skills for infographics and design to → tweet link
@Teknium · 2026-04-19T06:56
RT @ChuckSRQ: 🚀 Artifact Preview v3.0 just shipped for Hermes Agent! → tweet link
@Teknium · 2026-04-19T02:31
RT @retakedottv: meet Mira, our submission for the @NousResearch hackathon. she's an AI streamer generating DJ tracks live → tweet link
Developer Tools
@ollama · 2026-04-19T02:08
ollama launch copilot Ollama now supports GitHub's Copilot CLI, the terminal agent that works directly with repositories on GitHub. → tweet link
@ollama · 2026-04-19T03:44
RT @KevinCrosby1: Congratulations to the @ollama team for shipping the Copilot CLI support! → tweet link
@ollama · 2026-04-19T03:44
RT @BradGroux: This is awesome news, @ollama now supports the fantastic, @GitHub Copilot CLI! → tweet link
@ollama · 2026-04-19T02:42
RT @Saboo_Shubham_: This is HUGE. Ollama now supports Hermes AI Agents. → tweet link
@steipete · 2026-04-18T19:17
🎚️CodexBar 0.21 Abacus AI provider, Codex Pro $100 support, safer OpenAI web extras, fixed local cost scanning, z. ai 5h quotas, Antigravity/Cursor/Ollama fixes → tweet link
@steipete · 2026-04-18T19:19
RT @elonmusk: You can access 𝕏 APi via @OpenClaw. → tweet link
@steipete · 2026-04-18T19:18
RT @AdamLowisz: @elonmusk @openclaw Grok has gone OpenClaw, unleash the lobsters 🦞 → tweet link
@TheAhmadOsman · 2026-04-18T22:37
Agent / harness traceability Allows you to ask an agent to basically remind you of what happened in a previous session and cross-check across different agents / harnesses as well → tweet link
@TheAhmadOsman · 2026-04-18T19:57
Can LLMs do X? The answer is almost always yes provided that you've trained the LLM on enough data pertaining to X → tweet link
@badlogicgames · 2026-04-19T14:36
RT @vercel: We've identified a security incident that involved unauthorized access to certain internal Vercel systems → tweet link
@nummanali · 2026-04-19T14:26
Vercel has an Internal Security Incident 🚨 Unauthorized access to their internal system - this is very serious → tweet link
@nummanali · 2026-04-18T19:14
Don't forget that you can create a skill with: $ skill-creator (codex) /skill-creator (claude code) → tweet link
@TheAhmadOsman · 2026-04-19T00:29
Qwen3.5-9B on an RTX 3070 8GB (llama.cpp vs TurboQuant) - UD-IQ3_XXS → 262k ctx across the board 8GB cards can push stupid context lengths now → tweet link
@TheAhmadOsman · 2026-04-18T22:55
Benchmarked the same Qwen3.5 9B UD-IQ3_XXS GGUF on an RTX 3070 8GB using Latest upstream llama.cpp VS TheTom's TurboQuant llama.cpp fork → tweet link
@badlogicgames · 2026-04-19T12:45
RT @walterra: Wrote down some notes how I set up pi-coding-agent + local Qwen3.6-35b-a3b on a M4 Max 128GB. Tests so far: Gave me a nice b → tweet link
@badlogicgames · 2026-04-19T17:00
RT @anemll: Added docs for running the Flash-MoE server with mariozechner's pi-agent @badlogicgames It already looks pretty snappy on Mini → tweet link
@badlogicgames · 2026-04-19T16:14
RT @ivanfioravanti: oMLX is working really well as single machine inference engine for coding agents! → tweet link
@steipete · 2026-04-19T07:49
RT @liuliu: You cannot believe how much intrinsic knowledge GPT-5.4 has about building GPU compute kernels. → tweet link
@steipete · 2026-04-19T08:00
RT @HamelHusain: Lots of people asking what's so good about the new codex desktop computer use. Here's 5 things that come to mind → tweet link
@TheAhmadOsman · 2026-04-19T00:03
results for best settings all of this was automated by codex-cli btw → tweet link
@TheAhmadOsman · 2026-04-19T09:01
You can tell Codex Cli to build you a scanner tool for the printer you have connected to the WiFi network faster than trying to find a good one → tweet link
@gdb · 2026-04-18T20:05
incredibly fun to build webapps and games with codex, entirely with natural language → tweet link
@TrungTPhan · 2026-04-19T18:08
RT @bearlyai: This demo shows the Calude integration with Powerpoint and it's possible to auto-generate a "Mckinsey" look → tweet link
@RealGeneKim · 2026-04-19T18:01
RT @manicode: I'm gonna use Opus 4.7 and Codex 5.4 to do a full threat model and code review of my favorite Security library of all time. → tweet link
@steipete · 2026-04-19T18:16
RT @onusoz: Who is running local models on GPUs on OpenClaw? I have started benchmarking different models this week. → tweet link
Security & Infrastructure
@nummanali · 2026-04-19T14:26
Vercel has an Internal Security Incident 🚨 Unauthorized access to their internal system - this is very serious, and you should take action if your store secrets with them → tweet link
@badlogicgames · 2026-04-19T18:22
RT @zeeg: If you want to rotate your SENTRY_AUTH_TOKEN: Settings > Organization Tokens > Create New → tweet link
@hnasr · 2026-04-19T14:01
If you ever used the /+ APPEND / Oracle hint to bulk insert rows into a table, you won't be able to read the table in the same transaction. ORA-12838: cannot read/modify an object after modifying it in parallel. → tweet link
Platforms & Ecosystems
@victormustar · 2026-04-19T04:58
RT @ClementDelangue: Hugging Face is becoming the platform for agents to use and build AI. Now they can call 1M HF spaces to do everything → tweet link
@alexocheema · 2026-04-19T11:09
RT @0xSero: I interviewed Huggingface's head of product Victor. Huggingface is AI's core → tweet link
@steipete · 2026-04-19T14:45
RT @davemorin: The OpenClaw story on the @TEDTalks stage. @steipete knocked it out of the park. Standing ovation. → tweet link
@juliarturc · 2026-04-19T17:57
Oh no I haven't yet tried OpenClaw and it's already dead. Can I get an extension for a couple of weeks plz? 🥺 → tweet link
@nummanali · 2026-04-18T23:45
How is everyone sleeping on the OpenResponses spec It's based on the OpenAI Responses API and is backed by every major provider bar Gemini and Anthropic → tweet link
@RydMike · 2026-04-19T11:43
RT @tsuyoshi_chujo: GetX repo is now back #FlutterDev → tweet link
@victormustar · 2026-04-19T09:08
RT @songjunkr: 가장 좋아하는 로컬LLM 양자화/파인튜닝 리스트 → tweet link
Hardware & Local Inference
@alexocheema · 2026-04-19T15:11
RT @songjunkr: 애플이 6월 WWDC에서 맥스튜디오 M5 Ultra를 공개한다면, 한시간안에 예약주문이 수십만건 들어올거야. 대형 로컬LLM을 돌리기 가장 베스트거든. → tweet link
@sudoingX · 2026-04-19T05:21
i have tens of cases in my DMs. companies running sota models on enterprise clusters for workloads that could have been tested and validated on 4x RTX 3090s first. → tweet link
@TrungTPhan · 2026-04-19T14:54
A bit shocked by the winner of the Chinese humanoid robot marathon. → tweet link
@steipete · 2026-04-18T19:49
for me it's really just about winning on view counts. damn you @steipete → tweet link
AI Applications & Industry
@FinansowyUmysl · 2026-04-19T08:22
O kurczę. Nie sądziłem, że nastąpi to tak szybko. Myślałem, że zanim ludzie wsiądą do autonomicznego autobusu, wcześniej muszą upowszechnić się autonomiczne samochody. A tu proszę. W Norwegii będzie trasa z pierwszym autobusem bez kierowcy. → tweet link
@FinansowyUmysl · 2026-04-19T05:25
Bardzo ciekawy komentarz CEO Stripe o tym, że do analizy sekwencji genów wykorzystuje… agentów AI. Dzięki nimi dowiedział się, między innymi, że ma 30x większe ryzyko czerniaka niż średnia populacji. → tweet link
@badlogicgames · 2026-04-19T16:15
RT @DanielGri: So happy @badlogicgames open sourced https://t.co/5bP8QnZC8u - feels magical if you can unbrick a broken Magento shop checkout → tweet link
@nummanali · 2026-04-18T18:40
Beka will be an industry defining leader He's the creator of Better Auth → tweet link
@thdxr · 2026-04-19T06:46
RT @opencode: Good luck to everybody taking part in the OpenCode Buildathon today → tweet link
Developer Community & Culture
@LinusEkenstam · 2026-04-19T10:00
It's not just fringe, the early adopters into the analog wave, have been silently paying attention for a long while Cassettes, Vinyl, iPods — zines. It's all coming together Apples biggest missed opportunity this year is NOT making a 25th year iPod classic anniversary model → tweet link
@LinusEkenstam · 2026-04-18T20:24
I'm starting to mass produce these little fellas in my garage. Will soon be up on website for purchase. → tweet link
@thdxr · 2026-04-19T07:06
the issue every single company is having is they're looking at what they currently do and figuring out how to branch off into ai everything that comes out of this has been useless just think about what they'd build if starting today → tweet link
@thdxr · 2026-04-19T04:32
if they limited claude access to just people named claude they'd probably solve their compute issue overnight → tweet link
@kunchenguid · 2026-04-19T02:12
hard to find something more disappointing than: - having written a perfect prompt of a big task and sent to agent - went to dinner - came back seeing this agent: "just one more question before I start..." → tweet link
@sudoingX · 2026-04-19T05:59
the messenger of intellect (nous) pervades everywhere. → tweet link
@sudoingX · 2026-04-19T05:21
stop renting H100s by the hour before you know what you're solving. if you throw compute at a data problem you will burn through funding before product discovery. think sir the order matters more than the budget. → tweet link
@TheAhmadOsman · 2026-04-18T21:19
Modern-day dev - Pays $67/month for a tiny VM - Can't edit a file in nano - Terrified of SSH - "What's a port?" - Engineer → tweet link
@RydMike · 2026-04-19T01:27
RT @ylecun: Dario is wrong. He knows absolutely nothing about the effects of technological revolutions on the labor market. → tweet link
@thdxr · 2026-04-18T19:45
ai will destroy
% of entry level jobs in years → tweet link
@badlogicgames · 2026-04-18T19:58
RT @rachelnabors: Weekend mini-hack complete: now I have a little Claude pet who can hassle me when I'm needed in the loop 😈 → tweet link
@thdxr · 2026-04-19T18:16
one leg back to normal one to go acting like his old self → tweet link
Open Source Projects
@badlogicgames · 2026-04-19T18:20
this is a bit wordy, and the proposed methods are less reliable than they are being made out. but it's actually directionally correct and worth a read. → tweet link
@badlogicgames · 2026-04-19T17:18
my thoughts go out to the triangle company team. this sucks :/ → tweet link
@badlogicgames · 2026-04-19T12:47
RT @helloiamleonie: embracing the cringe until further notice → tweet link
@badlogicgames · 2026-04-19T09:47
RT @mattzcarey: love this from @timzaman → tweet link
@badlogicgames · 2026-04-18T19:39
RT @bentlegen: new version of slop-scan (0.3.0) is out: - 🆕
deltacommand to compare reports or repos - 🆕 plugin support → tweet link
@badlogicgames · 2026-04-18T19:37
RT @steipete: 🎚️CodexBar 0.21 → tweet link
@badlogicgames · 2026-04-19T08:54
RT @ethanolivertroy: Migrating this old golang project I had to browse the KEV database and map it with EPSS → tweet link
@badlogicgames · 2026-04-19T06:10
recommended reading. → tweet link
@badlogicgames · 2026-04-18T19:56
recommended viewing. specifically because cristina is awesome. first ever talk is a keynote in front of a huge audience and she didn't even flinch. → tweet link
Miscellaneous Tech
@MatejKnopp · 2026-04-19T09:23
Your 1s and 0s will never sound more identical than with this 1445.63 EUR ethernet cable. → tweet link
@ASalvadorini · 2026-04-19T03:07
RT @TahaTesser: Realtime object detection with #Apple Core ML → tweet link