Vlad Tansky (@vltansky) header
Vlad Tansky (@vltansky) profile picture

Software Engineer @Wix | Codex Ambassador - Tel Aviv 🇮🇱

Israel
Joined May 2012
Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)Photo by Vlad Tansky (@vltansky)
Everyone is talking about Codex, Claude Code, and coding agents, but I’m rediscovering ChatGPT during my current trip to Japan. Another small tip I recommend trying: When planning a route with ChatGPT, ask it to include Google Maps links. I tried this a year ago and it was really unreliable. Now it works great, and can even build a route with multiple stops and give you a single Google Maps link for the whole thing. Combined with the new Dynamic UI, it works amazingly well!
5
96
Vlad Tansky retweeted
powered by OpenAI plugin extensions 🤯 github.com/openai/mcp-extens…
Replying to @claudeai
Also new: with interactive connectors, you can now open the full version of a tool inside Claude. Work in Adobe, Atlassian, and Box beside your conversation. Make changes yourself, or ask Claude to make them with you. Rolling out on web and desktop.
21
25
7
383
82,788
Inspired by @nickbaumann_ post about giving your personal agent responsibility, I asked my Dot to manage our ongoing Japan trip. And I love how proactive it’s been! We wanted to take a scenic train ride, but tickets were sold out, so we gave up on it. Today, some extra seats opened up, and my Dot spotted them and notified me! And it’s not the first time it’s improved our trip without me even asking. Go read Nick’s post (link in comments) and give it a try!
3
4
150
How is no one talking about how good ChatGPT’s visualizations have become? It’s genuinely so much more enjoyable to consume its answers now. Image grids, icons, interactive maps, sliders, embedded YouTube videos (even inside slides!), and even using your current GPS location. I’ve been using it a lot during my Japan trip, and it’s been incredibly useful. Can’t wait for Dots to support the same output format. Walls of text suddenly feel so outdated.
1
6
105
@ajambrosino @thsottiaux @reach_vb bring this to dots please 🙏
2
46
Starting today, eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in API credits to build apps and agents.
643
807
967
9,201
2,136,606
Vlad Tansky retweeted
Inbox Zero Mail is live on Product Hunt! It's the open source Superhuman alternative. I'd love your support: producthunt.com/products/inb… ⚡ Fast and keyboard-first 📥 All your Gmail and Outlook accounts in one inbox 🗂️ Split inbox ✍️ AI drafts in your voice 🧹 Bulk unsubscribe 💬 Team comments 📱 Web, Mac, Windows and iPhone
91
58
17
260
139,013
Vlad Tansky retweeted
It's only been a year.
Today we're introducing Skills in claude dot ai, Claude Code, and the API. Skills let you package specialized knowledge into reusable capabilities that Claude loads on demand as agents tackle more complex tasks. Here's how they work and why they matter for the future of agents:
67
70
13
1,919
70,832
Vlad Tansky retweeted
Teaser for Octocode v20🐙 (probably the last major version I will publish) Well...a lot of people claim nowadays that the top tier models "knows what to do", and many are surprised when I tell them that I'm still working on Octocode. I'm working on improvements and efficiency for about a month now, and I'm happy that benchmarks shows my work better than words . Compared to the same 30 research questions, v20 used: ~𝟐𝐗 fewer characters than Octocode v18 ~𝟒𝐗 fewer than plain gh ~𝟖𝐗 fewer than gh + RTK ~𝟕𝐗 fewer than gh + Headroom ** The correctness level didn't change l hope I will be able to publish Octocode MCP v20 on the next week github.com/bgauryy/octocode/…
1
1
4
141
Vlad Tansky retweeted
we removed a toggle from chatgpt! no, not that one. but, you no longer need to toggle developer mode to connect custom MCP servers! all users can now just go to plugins → add → create custom MCP server to connect MCP servers.
146
41
28
1,118
91,907
Vlad Tansky retweeted
I have not seen near enough excitement about MCP Events developers.openai.com/plugin… Right now, the only way most people's agents can wake up and do something is either A) cron, or B) you decide to message it Event-driven triggers are a huge deal
124
121
35
1,723
306,382
Vlad Tansky retweeted
OpenAI is preparing a new higher-tier ChatGPT Pro plan
204
67
173
1,510
519,626
Recommended read. Don’t use Astra as your main model - it will burn your usage fast. Use it as advisor subagent
I had Astra analyze all my Codex threads and learned a lot about quota. Here's how to get the most Astra usage. (applicable to Fable too) I've spent ~$5000 on Astra since release*: - 70% is JUST cached input reads - 16% is uncached input - 13% is output - <1% is cache writes Till now, my Codex threads kept getting longer and longer, but Astra broke the trend (see chart). Long context is eating most of your quota. The biggest levers: - Smaller threads. New thread for each little task. Don't let context build up - Use smaller models that delegate coding tasks to Astra subagents. Don't let Astra see big outer thread - Avoid continuous polling. Have Astra set timers and wake back up for long-running tasks - Trim tokens wherever you can: skills/MCPs, AGENTS.md, custom instructions, etc - 3rd party tools like RTK that return more compact outputs might help (haven't tested this yet) Investigating that 70% input cost in more detail: - 40% is encrypted content (reasoning or other OpenAI stuff) - 20% is file reads - 10% is tool call arguments, schemas, etc - 7% is skills (catalog, loading instructions, output) - 5% is base instructions (system prompt, custom instructions, AGENTS.md) - 5% is browser/computer use input - 3% is command/tool output - The rest is minor or not actionable: compaction handoffs, assistant prose, user messages What you can do: - Use the lowest reasoning you can for your task, to cut CoT tokens - Make your repo as agent-friendly as possible. Small files/modules with clear folder structure, so Astra isn't grepping and reading large chunks. Try asking Astra to delegate file search tasks to subagents like Luna - Don't dump in tons of skills, instructions, AGENTS.md stuff; be intentional about it - Delegate browser/computer use to smaller models unless you need max performance And, probably goes without saying: - Don't use /goal with Astra. The few times I did absolutely obliterated my quota - /fast mode basically only if Tibo says he's resetting in an hour and you have 100% quota to burn Let me know if you have any observations to share! *I didn't literally drop $5k, this is mapping token to equivalent API prices
1
129
73
659
83
8,565
256,028
Just discovered this awesome feature in Codex: `⌃⇧M` lets you quickly jump between recent models I recommend mapping it to `CMD+.` Settings -> Keyboard shortcuts -> Open model picker P.S. I love Astra, but can’t afford it as my default, so I’ve been switching models a lot recently
1
6
169
Vlad Tansky retweeted
Introducing hi.new! A way for bots to talk 1:1 or in groups. No more copy paste between your bot and mine. They just talk. Find a time we're both free. Chase the invoice. Plan the offsite with the whole team's bots. - DMs and group chats - E2E encrypted. Open source - Works with Grok Bot, Rakazo, Hermes, OpenClaw - Only bots you approve can message yours - Free. Short names are paid Give your bot a name at hi.new. Your bot can even sign itself up and pay on its own with Link!
105
23
52
276
118,823
Vlad Tansky retweeted
This one is for 6-year-old me 🤯 Play infinite worlds in VR. Your actions shape the world, unscripted.
45
26
20
398
510,138