Anthropic's Claude Haiku 5.5 arrives beside GPT-6 Luna at $0.101 per million input tokens. On the same day, GitHub put it inside Copilot and Sonnet 5.5 cache reads dropped to that same rate.
Reused Sonnet 5.5 input now costs $0.101 per million tokens, the same as Haiku. Sonnet output still runs 20 times higher, so long responses are where agent margins drain.
Source: venturebeat.comDevelopers keep the coding agent they already like. OutSystems controls which systems that agent can reach, which helps teams whose real holdup is compliance sign-off.
Source: outsystems.comMistral's new one-trillion-parameter multimodal model is pitched as a European alternative to the leading closed and open AI systems.
Anthropic's Claude Haiku 5.5, out alongside GPT-6 Luna, costs $0.101 per million input tokens, a tenth of Haiku 4.5's $12.
Why it mattersBudgets built on Haiku 4.5's $12 input price now overstate spend for high-volume classification and support work, and GPT-6 Luna faces a far cheaper rival there.
Continue readingAnthropic also halved Sonnet 5.5 cache-read pricing to $0.101 per million tokens.
Russia's October 7 drone-and-missile attack on Ukrainian energy and infrastructure killed at least 24 people, including 18 in Pryluky, four of them children, after a missile hit a building.
Ether led the session's declines at -4.27% and Bitcoin fell 2.68%, both steeper than the 0.22% slips in the S&P 500 and Nasdaq, while the crypto Fear & Greed index still reads 64 (Greed).
A weekly operator deep dive, the full Dojo Pro prompt packs, and the complete prompt database. Founding members lock the launch price forever.
Join the founding list →The monthly share of firms switching AI model providers, averaged over three months, hit a record 8% in September. Anthropic's share rose after Opus 4 and Sonnet 4 launched; switching then cooled until OpenAI's June price cuts and new models revived it. The pattern ties provider churn to model launches and price moves.
Capcom is building REX, a major overhaul of its RE Engine pitched as a 'game engine of the AI generation.' Programmer Satoshi Ishida says AI will mainly streamline workflows that have slowed as games and assets grew. The company has said AI will assist with graphics, sound and programming.
FinCEN withdraws two unfinalized proposals: its 2023 plan to designate international crypto mixing a 'primary money laundering concern' under Section 311, and a 2020 rule that would have required banks and money services businesses to verify identities and keep records on self-hosted wallets. Neither was finalized, so no rule in force changes; the move takes two pending obligations off the table.
Cache First, Then Pick Your Model
Read full analysis
Tool of the day
5.713/ 10
OutSystems now lets Claude Code, Cursor and the other supported agents write app logic inside its governed platform, so the agent your developers like answers to the same permission rules as everyone else.
Agent Experience lets developers keep their favorite coding agent while OutSystems decides what that agent may touch, which suits shops where compliance sign-off is the real bottleneck. Organizations already standardized on OutSystems should pilot it, because the permission and compliance rules they built carry over to agent-written code. Newcomers have to absorb the whole platform before an agent writes a line, so it is a weak first stop.
Kiro Workflows is a workflow engine that sequences agents, runs parallel branches and loops, switches models mid-run, and adds test or review gates before work moves forward. Start by turning one pipeline you already run by hand, such as implement, test, then review, into a saved recipe, so the next project starts from a known-good template. This was reported by industry news sources, so confirm availability on kiro.dev before you build on it.
Try itTogether Link is a command-line tool that connects Claude Code, Claude Desktop, Codex, OpenCode, and Pi to Together AI's hosted models, and it can route each request to a model automatically. Use it to cut the cost of a coding agent you already rely on: point one low-stakes repo at an open model, compare output quality against your usual setup for a week, then decide which tasks can move over.
Try itThe Computer Use feature in GitHub Copilot CLI entered public preview on October 1. It lets Copilot operate desktop applications by clicking, typing, scrolling, and pressing keys. Point it at legacy software with no API, or at GUI-only test flows you currently click through manually. Because it is a preview, run it in a sandboxed account or virtual machine before you let it touch production systems.
Try itMove.ai uses patented AI software to extract high-fidelity motion data from video, so creators can animate characters and digital scenes at scale. Try filming a short movement sequence and importing the result into your 3D or game pipeline. If the quality holds up for your project, it can replace a session in a dedicated capture studio.
Try itDer tägliche Begleiter für agentic coding und Enterprise-Arbeit Prompt-Caching Lesen $0.20/ MTok Schreiben $5/ MTok Eingabe $4/ MTok Ausgabe $20/ MTok Sonnet 5.5.
claude.com Together Link beta. Together Link runs six coding agents on models hosted by Together AI: Claude Code, Codex, OpenCode, and Pi Code in the terminal, plus Claude Desktop (including Cowork) and ChatGPT Desktop.
docs.together.aiA search tool never says no: it returns its top-k passages even when the index holds no answer, so the agent sees irrelevant text instead of a miss signal. We ask what frozen search agents do when the tool refuses instead.
N-explain/O | 0.8 | 92.6 | =V | 0.7 | 95.0 | =V |Read the paper
Receipts come from the Firecrawl Developer Index (READMEs, docs, issues, merged PRs), first-party sources only. The paper comes from the Firecrawl Research Index: arXiv, indexed in the last 7 days and submitted this month or last, picked for today's lead. Quotes are verbatim.
Meter output tokens separately. Split your Sonnet 5.5 logs into cached input and output so you can see where spend actually lands now that reused input costs $0.10 per million tokens.
Pilot OutSystems Agent Experience. Connect the coding agent your team already uses. Write the permission list for what it may touch before the compliance review starts.
Trial Haiku 5.5 in GitHub Copilot. Set the model picker to Claude Haiku 5.5 for routine edits for one week, then compare acceptance rates against your current default.
Haiku 5.5 input and cached Sonnet 5.5 input both cost $0.101 per million tokens, so the input side of an agent bill barely separates the two models anymore. The gap that remains is output, which the Cache First deep dive puts at 20 times more on Sonnet. Cache every reused prompt prefix and cap response lengths on Sonnet calls this week. Provider switching hit a record in September, which suggests competitors are already rerouting work to wherever these new prices fall.
Source: venturebeat.comMethod names the computation path behind the row, from a closed vocabulary: as published, yfinance close, 03B search corroboration, 04R dossier scoring, not recorded.
A figure is money, a percentage, a spelled magnitude, a multiple, a rate carrying its unit, a score over its denominator, a thousands separated number or an index close. Bare integers and years are not counted, so a phrase like "over 100 integrations" carries no mark. A chart's own scale line is the axis of the level above it and is counted once, with that level.
195 editions so far, one a day. Unsubscribe anytime.
Forward this to one operator you work with. Your referral link is in every email; milestones at koda.community/refer.