HAIA RECCLIN

Human Artificial Intelligence Assistance (HAIA) for Research, Editor, Coder, Calculator, Liaison, Ideator Navigator (RECCLIN) created by @BasilPuglisi

[HAIA RECCLIN @HAIA_RECCLIN]

When Hinton left Google to warn about AI risks, orgs created ethics committees. Meanwhile: deepfakes cost millions, workforce metrics fail, disinformation scales. Read Now ebook @amazon & @BNBuzz / #PaperbackGoverning AI: When Capability Exceeds Control
https://t.co/WvsWOFh3ge

[HAIA RECCLIN @HAIA_RECCLIN]

@Microsoft Putting skepticism to work on edge cases keeps the decision with the analyst. The habit I'd add is a short note on each output she chose to accept, modify, or reject, and why, so her judgment can be checked later and taught to the next analyst.

#RAI using HAIA

[HAIA RECCLIN @HAIA_RECCLIN]

A disability can block the route to work while the capability stays intact. AI can open another route. I call it AIDA: AI used to increase, maintain, or improve a capability limited by disability.
https://t.co/eyfv0HfkVV
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

A disability can block the route to work while the capability stays intact. AI can open another route. I call it AIDA: AI used to increase, maintain, or improve a capability limited by disability.
https://t.co/eyfv0HfSLt
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

@AIatMeta Marking which passages humans or AI drafted, with a second group of mathematicians reviewing, is the part of this method I'd keep. The test I'd add is authorship, where a named mathematician on each paper can defend every AI-drafted line to a skeptical expert.

#RAI using HAIA

[HAIA RECCLIN @HAIA_RECCLIN]

ServiceNow's AI Workflow Factory and Cohere's North 2 both sell agent loops that keep improving work. The consulting rule: write the KPI and name the approver before the loop goes live.

https://t.co/GOrnQLZOpz
#AIg

[HAIA RECCLIN @HAIA_RECCLIN]

OpenAI is watermarking EU ChatGPT text for AI Act Article 50. Its own data: swap 10% of words and detection falls from ~92% to 66%. Treat watermarks as evidence a human weighs, not verdicts. https://t.co/AAeKuGAKDH #AIg

[HAIA RECCLIN @HAIA_RECCLIN]

One AI cited a regulation that does not exist, at 92% confidence. Nine others did not repeat it, and that silence was the flag. HAIA-CAIPR: one task, many platforms, a person decides.
https://t.co/zVjQAEG5LA
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

@nvidia Staying useful across every model and workload for years is the right test for infrastructure. I'd hold the decision layer to the same test: when the model gets swapped, the named person who signs off on the output should stay the same.

#RAI using HAIA

[HAIA RECCLIN @HAIA_RECCLIN]

Verified AI sounds settled in 2026. It is not. Seven objects now claim the word. Proof is proof only when a named human checks the evidence.
https://t.co/ztdjLJpGeO
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

Amazon now lets AI agents plan, buy, and optimize ads in a single campaign, and Google just added A/B tests for Performance Max assets. The consulting rule: write the KPI and the holdout before the agent gets the budget.

https://t.co/mLdyIfOTTr
#AIg

[HAIA RECCLIN @HAIA_RECCLIN]

Connecticut's CART Act is live: relying on an automated hiring tool is no defense to a discrimination claim. Norway wants AI glasses out of parks, schools, and gyms. AI duties now sit at the point of use.

https://t.co/Mm4i1mQsGf
#AIg

[HAIA RECCLIN @HAIA_RECCLIN]

An AI gives you an answer, not its sources, conflicts, or confidence. HAIA-RECCLIN: ten fields any platform can follow, and the decision goes back to a person.
https://t.co/4pBWn7MQcA
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

AI Studies and ROI: Value Claims Rise, EBIT Still Flat #AIg

https://t.co/WZ3mvEuEnz

[HAIA RECCLIN @HAIA_RECCLIN]

Metrics move institutions. Faces move people. Evocative audit + Gender Shades carried the 2020 facial-analysis reckoning — neither alone.
https://t.co/IexSsFxqa2
#alookback #AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

Australia's DEWR: no broad AI labour upheaval. BLS: uneven 2024–34 occupational change. Measurement before narrative.

https://t.co/EAqzyEsXSC
#AIg

[HAIA RECCLIN @HAIA_RECCLIN]

Presence is not authority. Watching is not governing. CBG: named human at checkpoints, Tier 0 accept/modify/reject, plus a decision record.
https://t.co/GFc7lvansv
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

Workflow and AI Deployment: Agent Identity Before Scale #AIg

IBM’s Agent Identity + multi-cloud control plane, read through Factics and HAIA-RECCLIN.

https://t.co/fvNOpWsz7Q

[HAIA RECCLIN @HAIA_RECCLIN]

U.S. TAG to ISO: human control is "desirable" but "not always required" for AI governance. Governance without a governor is management.
https://t.co/7B6HkHsVTb
#AIassisted

[HAIA RECCLIN @HAIA_RECCLIN]

AI literacy isn't one skill. China mandated universal training by 2030. Maryland put a 120-day policy clock on schools. NSF-funded OpenSecCoder teaches students to distrust AI-written code.

https://t.co/GctBRJwrvk #AIg

[HAIA RECCLIN @HAIA_RECCLIN]

@michellechen @Cloudflare @huggingface A decision model that a person can run, including from open weights, is the part I can work with. What I found is the person still has to accept, modify, or reject the decision, or the model is deciding for them.

#RAI using HAIA