
@MrTechnologyXi
iAccount based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States Android App
Account-level information from X, not a live location or the device used for a specific post.
Joined October 2025
- Tweets58
- Following32
- Followers3
- Likes17
OpenAI API usage tiers went from five to three.
Build: $5 in total payments
Launch: $100
Grow: $500 (the top tier used to need $1,000)
Upgrades are automatic. Nothing to do.
developers.openai.com/api/do…
#LLM #OpenAI
After ArtCraft, the creator is now going for Free, open-source Word, Excel and PowerPoint clones: WordCraft, GridCraft, DeckCraft.
Claude Opus 5.5 is a commit co-author.
All alpha, 50-60% ready and what's left in agent-hours.
github.com/storytold/craft-l…
#AIAgents #OpenSource
How to read an AI math claim:
OpenAI posted 722 manuscripts (372 result families) from an unreleased model.
Lean-checked: a proof checker verified it.
Not formalized: no machine check.
Peer reviewed: none yet.
Critics flagged missing disclosures.
github.com/openai/math
Harness: tool missing, can’t read the file, permission denied, limit hit.
That’s the wrapper. Change what’s wired in, not the model.
Agent: plans forever, same step again, drifts off the goal.
That’s the loop. Change the goal, the plan, or when it stops — not the app.
Most people call the whole app an agent.
That’s two things mashed together.
An agent is the model loop — it plans, calls tools, iterates toward a goal.
A harness is the app or CLI around that loop — tools, files, permissions, rate and spend limits.
When something breaks, naming the layer matters.
Blame the model when the harness is blocked, and you pull the wrong lever.
Swap harnesses every time the loop hiccups, and you never finish.
Pick one harness you can finish a real task in.
Stay with it until that task ships.
#AIAgents #LLM #AI
SMH!
Tomorrow we will bring back the 5h limit for Plus accounts across ChatGPT Work and Codex. I had mentioned this a while ago, but then postponed it.
This is necessary as (a) the 5h limit allows us to smoothen the load on our compute, allowing to keep the plan generous in terms of weekly usage and (b) users on the Plus plan are relatively casual and new users, but then also just accidentally eat through their whole weeks usage and then are confused, making it not a great experience.
We are for the upcoming months keeping the 5h limit not enabled for Pro $100 and Pro $200 subscriptions.
Open-weights cut my $34K/month API bill to $3.1K. Crossover is 60-200M tokens/mo. Stack: DeepSeek V4-Pro + Qwen 3.8 + GLM 5 + Muse Glimmer + LFM2.5. mr.technology/payloads/self-… #AI #LLM #OpenWeights #SelfHosted #AIAgents
mr.technology retweeted
Use Hermes Agent however you want.
Modify it, extend it, share it, fork it. Use it locally or in the cloud. Use the desktop app or the TUI or text it or talk to it from your smart fridge. Set up your great-aunt with it or run a Fortune 500 on it. Give it your own keys, use it through Nous Portal, or use local models on your own stack. Use one model or fifteen or blend them with MoA. Use it to plan an elopement, debug an EVM event, summarize a lease, troubleshoot that pesky issue with your car, reverse a DataFrame, or autorespond to your boss's emails. Run it on a Mac Studio, a $35 SBC, a hand-me-down ThinkPad, a Docker container, a Hermes Cloud VPS that costs as much as a cup of coffee, or eight of each at once. Create a shared bot that answers questions no one on your team can. Ask it to modify itself. Use it to parse a WhatsApp backup, answer your mother's questions about a power bill, or set a cron to scan the group chat and notify you when someone mentions that deadline you're avoiding. Put it in charge of remembering why you named the branch that. Put it in charge of filing your email into folders or cleaning up all the files named report_final_draft1.pdf. Have it scan the news weekly to settle that stupid bet you made with your brother that one time. Ask it to send your wife an article about cats every morning which she will pretend to be annoyed by even though it really makes her day.
Choose your own
Stripe paid $7B for OpenRouter — and now bills AI providers AND routes their traffic. Half the routed tokens already go to Chinese models.
mr.technology/payloads/strip…
#AI #AIAgents #OpenRouter #Stripe #LLM
Everyday Hermes Agent gets better
Your AI agent authenticates as you. The API can't tell it's an agent. Compromise it and every credential it ever touched walks out. mr.technology/payloads/ai-ag…
#AI #AIAgents #Security #MCP #ZeroTrust
HITL approve/deny prompts just failed at 33% across 40,000 plays. npm run analyze exfil approved 64.7% — payload visible in the log. mr.technology/payloads/agent… #AI #AIAgents #Security #ClaudeCode
6 vendors signed one plugin manifest. Amazon, Google, Microsoft, OpenAI, Cursor & Vercel agreed on Agent Plugins 1.0 — a two-line JSON schema and one directory layout. Already works in Cursor, Claude Code & VS Code. mr.technology/payloads/agent… #AI #MCP #DevTools #Anthropic
mr.technology retweeted
Agents should use tools that excel at what’s important for an AI model. Kitesurf is Cloudflare’s new stateless, highly scalable, and cost-effective web browser that runs entirely on top of Workers and was designed specifically for the Agentic Cloud. cfl.re/4fJnGGc #AgentsWeek
Competition is a great thing
We are committed to pushing the model frontier across cost efficiency, capability, and speed.
Starting today, we are reducing prices for GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20% , and offering a faster option for GPT-5.6 Sol in the API.
Luna and Terra’s lower prices are reflected in how usage is counted in Codex and ChatGPT Work, so your usage goes further.

















